The Relationship Between
the Growth of Health
Amenities and Human
Capital
THESIS WITHIN: Economics NUMBER OF CREDITS: 30 Credits PROGRAMME OF STUDY: Civilekonom AUTHOR: Johan Nyqvist
TUTORS: Agostino Manduchi, Aleksandar Petreski JÖNKÖPING May 2020
A study examining the relationship between the growth
of health amenities and the share of high human capital
individuals in the Swedish municipalities
i
Acknowledgments
I would like to take this opportunity to show my gratitude to my tutors, Agostino Manduchi and Aleksandar Petreski. I am sincerely thankful for the support, guidance, and valuable inputs they have provided me with throughout the writing of this thesis.
I would also like to express my gratitude towards Orsa Kekezi who assisted me in the selection of research topic and provided me with critical feedback on multiple occasions during the writing of this thesis.
Finally, I would also like to thank the participants of my seminar group for helpful comments and suggestions. Their inputs have been very valuable as they helped me understand the reader’s point of view of this paper.
ii
Master Thesis in Economics
Title: The Relationship Between the Growth of Health Amenities and Human Capital – A study examining the relationship between the growth of health amenities and the share of high human capital individuals in the Swedish municipalities
Author: Johan Nyqvist Tutor: Agostino Manduchi
Aleksandar Petreski Date: 2020-05-18
Key Terms: Health amenities, Human capital, Lifestyle, Health, Amenities, Sport amenities, Retail-oriented health amenities, Creative class
Abstract
Despite the well-known benefits associated with physical activity, the lack of physical exercise has risen to become one of the most severe health-related problems of the 21st century. Previous studies have found a good health profile and frequent engagement in physical activities to be positively correlated with productivity and mental well-being. Thus, it is of interest for the municipalities to better the health of their inhabitants. The accessibility of health-encouraging facilities, goods, and services have been discovered to be positively related to health and physical exercise. Hence, it is of interest to the municipalities to understand what drives the growth of these amenities. Especially considering that it is easier for municipalities to affect the accessibility of health amenities on a municipal level than changing lifestyle habits on an individual level.
The purpose of this paper is to examine whether the growth of “health amenities”, defined as amenities encouraging a healthy lifestyle, is related to the share of high human capital individuals in the Swedish municipalities. To examine this relationship, this paper inspects the percentage change of employees within these establishments across the Swedish municipalities between 2010-2019 through a series of OLS regressions. The results indicate that the growth of health amenities is positively correlated to the share of high human capital individuals in the Swedish municipalities which confirms the hypothesis of this paper.
iii
Table of Contents
1. Introduction
1.1 Purpose
2. Literature Review
2.1 Human Capital and Regional Development
2.2 Amenities and the Geographical Distribution of Human Capital 2.3 Environmental Characteristics and Physical Activity Levels 2.4 Educational Attainment, Socioeconomic Status, and Health
3. Hypothesis and Expected Results
4. Methodology 4.1 Empirical Model 4.2 Variables 5. Empirical Results 5.1 Descriptive Statistics 5.2 Correlation Analysis 5.3 Regression Analysis
6. Discussion and Analysis
6.1 Limitations
7. Conclusion
8. List of Reference
9. Appendices
List of Tables
Table 1: Variables, definitions, and expected signs
Table 2: Municipalities with the highest positive and negative percentage changes of employees in health amenities
Table 3: Descriptive Statistics Table 4: Regression results
1
1. Introduction
The relationship between educational attainment, socioeconomic status (SES), and health is well documented and there is a wide span of research linking them together (Florida, 2002b; Ford et al., 1991; Ross & Wu, 1995). High human capital individuals (Individuals with three years or more of higher education) have been discovered to value their health higher as they, in general, are more physically active and tend to have healthier nutrition habits than those of a lower human capital stock (Folkhälsomyndigheten, 2019b; Livsmedelsverket, 2016).
Health is a factor of multiple variables and to reach and maintain a good health, preventative measures are just as important as the medical factors (Folkhälsomyndigheten, 2019b). Regular physical activity has been recognized as a key factor of health as it is known to prevent obesity and an almost countless number of diseases (Cadilhac et al., 2011; Folkhälsomyndigheten, 2019b). In addition to preventing workers from getting sick, frequent engagement in physical activities has been found to positively influence mental well-being and productivity. This has led to physical activity levels and health being considered as important components of economic growth (Berger et al., 2001; Cadilhac et al., 2011). Notwithstanding the well-known benefits associated with physical activity, the problem of physical inactivity remains in the Swedish municipalities as well as in a global context (Folkhälsomyndigheten, 2019b), and Blair (2009) denoted physical inactivity as the greatest health-related challenge in modern times.
To address this problem, many researchers have examined how various environmental factors influence physical activity levels and health (Carr et al., 2011; Lee, 2019). The findings from these studies indicate that health and physical activity levels are positively correlated with the accessibility of health-related facilities, amenities, and socioeconomic status (Carr et al., 2011; Humpel et al., 2002). Macintyre (2007) suggested that the difference in physical activity levels between the socioeconomic classes could be due to what he refers to as deprivation amplification, which is the difference in the supply of health-oriented amenities between areas of different socioeconomic classes. Findings on whether deprivation amplification was present or not were scattered and inconsistent
2
(Giles-Corti & Donovan, 2002; Powell et al., 2006; van Lenthe et al., 2005). Hence, Macintyre et al. (2008) emphasized that policies must be developed from contemporary and context-specific empirical findings on the distribution of health-related resources. Kawakami et al. (2011) examined the relationship between neighborhood deprivation level and the availability of health-related resources in a Swedish context and discovered low SES neighborhoods to have better accessibility to health-related resources, goods, and services. However, since Kawakami et al. conducted their study in 2011, health-related industries in Sweden have skyrocketed. One example of the dramatic change can be seen in the gym industry, as the number of fitness facilities increased by approximately 90% between the years 2011-2016. The greatest increase in the number of fitness facilities was seen in the areas of Stockholm, Gothenburg, and Malmö and the difference in health correlates with the difference in SES across the Swedish municipalities (Folkhälsomyndigheten, 2019b; Mellander, 2019). These municipalities are the homes of most Swedes of a high human capital level (Mellander, 2016). In addition to putting a high value on their health, individuals of a high human capital stock have been found to value amenities higher than those of a lower educational attainment (Brueckner et al., 1999; Moretti, 2003). Furthermore, they have also been discovered to fuel the growth of amenities in the cities they inhabit (Shapiro, 2006).
The findings suggesting that high human capital individuals are drivers of the growth of amenities, along with their preferences for a healthy lifestyle, the correlation between health and socioeconomic status in the Swedish municipalities, and the explosion of fitness facilities in the largest cities in Sweden motivated me to explore the relationship between human capital levels and the growth of health-oriented amenities. This brings us to the main research question of this paper, whether there is a relationship between the growth of health amenities, defined as amenities encouraging a healthy lifestyle, and the fraction of high human capital individuals in the Swedish municipalities. This paper also examines if the growth of health amenities differs between the sport amenities (health amenities associated with physical activities) and retail-oriented health amenities (health amenities linked with retail activities of goods and services associated with a healthy lifestyle). Furthermore, it also compares how the growth of the different kinds of health amenities differs across the conventional and creative class measure of human capital.
3
Since the availability of these amenities is linked with the health and physical activity level of an area, which itself is connected with the region’s productivity level, it is important to understand what drives the growth of these amenities. Furthermore, studying the underlying factors to the growth of these amenities is appealing as it is easier for municipalities to impact the availability of health amenities on a municipal level rather than changing health-related behavior at an individual level.
Previous research on health-related resources has mainly studied the relationship between neighborhood deprivation level and the accessibility of health-oriented amenities (Giles-Corti & Donovan, 2002; Kawakami et al., 2011), the relationship between the accessibility of these amenities and physical activity levels (Carr et al., 2011; Humpel et al., 2002), and the benefits of physical activity (Berger et al., 2001; Cadilhac et al., 2011), whereas researchers examining the relationship between the growth of amenities and human capital levels mainly have focused on consumer amenities such as restaurants and nightclubs (Shapiro, 2006). To my knowledge, there is no academic research examining the growth of health-related resources. Furthermore, I could not find any previous research using “health amenities” as a joint term for amenities linked with a healthy lifestyle. These traits make my paper unique and I hope to contribute to the amenity literature by introducing the term health amenities and examining whether the share of high human capital individuals is related to the growth of health amenities in the Swedish municipalities.
4
1.1 Purpose
The purpose of this paper is to investigate whether the growth of health amenities is related to the share of high human capital individuals in the Swedish municipalities. To examine this relationship, this paper inspects the percentage change of employees within these establishments across the Swedish municipalities between 2010-2019 through a series of OLS regressions.
Furthermore, this paper also inspects whether the relationship between the human capital level and the growth of health amenities differs between sport amenities and retail-oriented health amenities. It also compares how the growth of the dependent variable differs across the conventional and creative class measure of human capital.
This thesis contributes to contemporary amenity literature by introducing the term “health amenities” and filling the gap on whether the human capital level is related to the growth of health-oriented amenities in a Swedish context.
5
2. Literature Review
The literature review provides the reader with a background of theories and previous findings laying the foundations for this paper. This section is divided into four separate parts. The first part explains the relationship between human capital and regional development. Subsequently, the second part discusses the geographical distribution of high human capital individuals and the underlying factors causing this distribution. The third part covers the relationship between environmental characteristics and physical activity levels. The final part of this section provides an overview of how health and lifestyle habits differ across education and socioeconomic status.
2.1 Human Capital and Regional Development
Numerous economists have studied the relationship between human capital and regional growth. Glaeser (1998), Glaeser et al. (1995), Glendon (1999) and Simon (1998) are just a fraction of those finding empirical evidence of a significant positive relationship between the two, and the fact that human capital plays a vital role in regional development is now well-established (Mellander & Florida, 2007). Despite the consensual agreement that human capital drives regional growth, there is a debate on the measurement itself. Meanwhile, the classic measure of human capital is considered as the share of the population with a bachelor’s degree or more, Florida (2002c) suggested that the measurement should be based on what people do rather than what they study. Thus, he introduced his own measure of human capital which he referred to as creative capital. He denoted the holders of this kind of capital as “the creative class”. These individuals are typically bohemians, artists, and gay people, working in creative industries. Mellander and Florida (2007) found the creative class measure to outperform the conventional human capital measure in a Swedish context, whereas Hansen (2007) found a high correlation between Florida’s super creative core and human capital, and concluded educational attainment to be an acceptable proxy for creative capital in a Swedish context. Population density and high human capital individuals are both well-recognized components of regional development and have been found to positively affect regional growth in many ways. Glaeser (2011), and Glaeser and Gottlieb (2006) discovered that large and dense cities with well-educated citizens had a comparative advantage in the
6
distribution of knowledge. Furthermore, Glaeser and Gottlieb (2006), along with Lucas (1988) and Romer (1989) acknowledged the importance of the spillover effects that comes as a result of density and a high frequency of face-to-face interactions. Glaeser (2011), and Glaeser and Gottlieb (2006) argued that the externalities generated by these spillover effects increased with the level of human capital as a higher level of high human capital individuals implies a higher possibility of talent teaching talent. Furthermore, Glaeser et al. (2001) discovered the population of dense areas to grow at higher rates in comparison to more sprawled areas.
Glaeser (2011) and Shapiro (2006) found highly educated individuals to be significantly more productive than those of a lower human capital stock. Additionally, a wide range of economic research has found the human capital level of a region to be related to the employment opportunities within the given area. Glendon (1999) and Simon (1998) discovered a significant positive relationship between the level of human capital and employment opportunities within a region, and Florida (2002a), Glaeser (1998, 2011) and Kotkin (2001) are few among those who argued that a high concentration and availability of skilled individuals could influence the geographical preferences of high-technology industries. They discovered that these industries were more likely to locate within areas with high human capital levels, and Florida (2002b) suggested that high-technology firms and human capital together are two independent key variables working together to increase regional incomes.
Florida (2002a) also recognized that a density of creative individuals of a high human capital stock was of great importance as he considered them to be natural producers of amenities, which have become increasingly recognized as a variable influencing the location decision of well-educated individuals (Gottlieb, 1995; Roback, 1982, 1988). Florida based this argument on findings suggesting that amenities grew at higher rates in regions with a higher number of creative and bohemian individuals. In line with the argument presented by Florida, Shapiro (2006) also suggested that individuals of a high human capital stock spur the growth of amenities in the cities they inhabit. He claimed that a high presence of these individuals generates a positive spiral effect of regional economic growth as the increase in amenities levels creates job opportunities and elevates the quality of life standards within the region. Subsequently, he claimed that this higher quality of life attracts more high human capital individuals, resulting in a concentration
7
of human capital and lifestyle amenities, which he argued would attract high-technology firms and boost the regional growth to higher levels in the long-run. This circle argument was later supported by Kahn (2008). He reasoned that high human capital individuals in general were higher earners than those of a lower education, hence they had more affluent income to spend on amenities. Kahn argued that the extra amounts spent on amenities would then attract amenities of a higher quality, which would then continue to draw more well-educated individuals, and so the circle goes on.
2.2 Amenities and the Geographical Distribution of Human Capital
In recent decades, well-educated individuals have increasingly been migrating to large cities around the world and we are now seeing an increasing accumulation of high human capital individuals in major cities (Baker, 2016; Berry & Glaeser, 2005; Florida, 2002b). This trend has been present in Sweden as well. Today, approximately 85% of the Swedish population reside in regions classified as urban areas, and more than 60% of the well-educated Swedes live in the three largest cities of the country (Mellander, 2016; Mellander & Bjerke, 2017). This distribution has not occurred randomly. It is rather the result of high human capital individuals self-sorting into urban municipalities with a high presence of amenities, job opportunities, and other factors matching their preferences (Baker, 2016; Moretti, 2003). While the significance of human capital to economic growth is consensually agreed upon there is still a debate regarding the factors that influence the location decision and yield the geographical distribution of talented individuals (Mellander & Florida, 2007).
The traditional theory of urban growth considers the geographical distribution of well-educated individuals as a function of production amenities such as thick labor markets, employment opportunities, and financial incentives offered by the cities. Despite the emphasis on the production side, the proponents of this theory recognize that consumer amenities play a complementary role in this distribution (Brown & Scott, 2012; Glaeser et al., 2001; Glaeser, 2011). Clark et al. (2002) argue that the classic view is becoming outdated in conjunction with globalization and other societal changes. The modern theories instead assign this allocation to a range of lifestyle factors such as consumer amenities and view the city as an entertainment machine (Clark, 2004). These factors have become central pillars of the so-called new growth theory and are increasingly being
8
taken into consideration in the decision making of policymakers (Mathur & Stein, 2005). This view is based upon the works of Jacobs (1961, 1969) who underlined the importance for cities to attract talent and promote diversity. Subsequently, vast economic research has connected amenities and lifestyle factors to regional success and the distribution of human capital.
Roback (1982, 1988) noted that quality of life amenities increasingly influenced migration decisions of talented individuals, and several economists suggest that cities can attract talent through offering a wide range of lifestyle amenities (Florida, 2000; Glaeser et al., 2001; Gottlieb, 1995; Kotkin, 2001). Brueckner et al. (1999) and Moretti (2003) found individuals of a higher educational attainment more likely to value amenities than those of a lower educational level. Florida (2002c) and Gottlieb (1995) suggested that this could be due to the fact that high human capital individuals in general have more affluent income than those of a lower education level. Meanwhile, Clark et al. (2002) argued that high-skilled individuals have a wider selection of cities to choose from than those of a lower education since they are offered more job opportunities. Thus, they locate within the area where the combined utility of job and leisure time is the highest.
Glaeser et al. (2001) argued that the attractiveness of an area is becoming increasingly dependent upon quality of life factors as the average person is getting wealthier. They also found that urban rents have increased at a higher rate than urban wages. This implies that the residents of these areas are financially worse off than those of more rural areas, suggesting that people migrate to urban areas for lifestyle habits and quality of life factors rather than for economic factors. Bjerke and Mellander (2017) discovered this to be the case in a Swedish context as well as the rent-premiums and extra costs associated with living in the larger cities in Sweden exceed the higher wages paid within these areas. Chen & Rosenthal (2008) found that preferences for amenities and migration patterns differ across the life-cycle, suggesting that age and stage of life are related to location decisions. The age to migration relationship is applicable in a Swedish context as well as 90% of the individuals migrating in Sweden are between 18-40 years old (Bjerke & Mellander, 2017). A large share of the Swedish migration can be connected to university studies. Mellander & Florida (2007) found universities being the most significant determinant in the location decisions of high human capital individuals in Sweden. In
9
addition to universities, Mellander and Florida also emphasized that a wide span of amenities, openness, and tolerance also influence the geographical allocation of human capital. They argued that these factors play complementary roles in the location decision of talented individuals as they appeal to different types of human capital.
In line with the findings of Mellander and Florida, Kulu et al. (2018) connected the increasing migration among young Swedish adults to the expanding post-secondary education. Furthermore, Hansen and Niedomysl (2009) found the majority of the relocation of the creative class individuals in Sweden to take place in conjunction with finished university studies. Gottlieb and Joseph (2006) examined the migration decisions of college graduates and learned that these individuals tend to migrate to areas with high levels of human capital. This could be due to the fact that well-educated individuals consider skilled neighbors as an attractive consumption amenity (Glaeser & Saiz, 2004). Clark (2004) suggested that university graduates are more receptive to constructed amenities (fitness facilities, libraries, museums, etc.) in comparison to natural amenities (lakes, temperature, weather, etc.) as he found them more likely to migrate to areas with a high concentration of constructed amenities.
Bjerke and Mellander (2017) studied the location decision of university graduates in a Swedish context. They found that most Swedish graduates choose to migrate to or remain in urban areas across the country. Hansen and Niedomysl (2009) argued that college graduates and human capital migrated for job opportunities rather than the quality of place. In subsequent work, Niedomysl and Hansen (2010) conducted a large-scale survey to examine the relative significance of jobs versus amenities for the decision to relocate for the Swedish migrants. The survey concluded that jobs are a substantially more important driver for the decision to migrate, but that outdoor activities and recreational amenities also mattered in the decision, especially among the Swedes of higher educational attainment.
10
2.3 Environmental Characteristics and Physical Activity Levels
Health is a factor of multiple variables and to achieve a good health and physical well-being, preventive efforts such as exercise and healthy dietary plans are of the same prominence as the medically oriented work. Physical exercise is widely recognized as a prominent deterring measure of a poor health profile as it among other things has been found to prevent obesity, diabetes, and cardiovascular diseases. Furthermore, it has been discovered to positively influence mental well-being which enhances overall health and productivity (Cadilhac et al., 2011). According to Kohl et al. (2012), physical inactivity was the fourth most frequent cause of death on a global scale in 2012, and Blair (2009) stated that the lack of physical exercise has risen to become the most severe health-related problem of the 21st century.
To address this problem, researchers have started to examine the relationship between environmental characteristics and physical exercise (Carr et al., 2011; Lee, 2019). Humpel et al. (2002) found a significant positive correlation between the accessibility of facilities and the physical activity level of an area. Carr et al. (2011) examined the effect of neighborhood walkability on health and physical wellbeing. Walking is the most commonly performed form of physical exercise among adults and previous findings suggested that a lack of walkable destinations was negatively correlated with physical activity levels (Owen et al., 2004). Similarly, Carr et al. (2011) discovered the availability of walkable destinations to be positively associated with physical activity.
A wide span of research within the area of environmental influences on physical activity has focused on the socioeconomic status and deprivation-level of neighborhoods (Kawakami et al., 2011; Macintyre, 2007; Powell et al., 2006). Typically, living in deprived neighborhoods of a low SES is linked with poor health status and high levels of smoking, obesity, and physical inactivity (Sundquist et al., 1999). Macintyre (2007), proposed that this could be due to what he referred to as deprivation amplification, suggesting that the low rates of engagement in physical activities in low SES neighborhoods may be due to these neighborhoods having lower accessibility of resources encouraging a healthy lifestyle. However, the relationship between neighborhood SES and physical activity have been examined by a wide range of economists and the findings have been rather divergent. Powell et al. (2006) found higher
11
SES neighborhoods in the US to have a wider range of physical activity centers available, while Giles-Corti and Donovan (2002) in an Australian context discovered affluent neighborhoods to have a higher access to golf courses and beaches, whereas the lower SES areas had a higher availability of gym facilities and recreational centers. Moreover, findings from the Netherlands indicated that the availability of health-related facilities did not differ significantly between low and high SES neighborhoods (van Lenthe et al., 2005). Recognizing the inconsistency of these findings, Macintyre et al. (2008) concluded that policies should be founded upon contemporary, context-specific data.
Kawakami et al. (2011) compared the accessibility to health-related resources between affluent and deprived neighborhoods in a Swedish context. Surprisingly, the results of their study did not align with the theory of deprivation amplification with all the investigated variables being more accessible in neighborhoods of higher deprivation levels. Although, the results also indicated that these areas had higher access to health-damaging goods and services such as fast-food restaurants and bars.
However, since Kawakami et al. conducted this study in 2011, the Swedish fitness industry has changed dramatically. An example of this striking change can be seen in the Swedish gym industry. Mellander (2019) found that the number of fitness facilities increased by approximately 90% Between the years 2011-2016. The municipalities seeing the greatest change in the number of fitness facilities across this time-period were Stockholm, Gothenburg, and Malmö. These are all large city municipalities of high socioeconomic status with well-educated populations, and more than half of the Swedes classified as high human capital individuals reside within these areas (Mellander, 2016, 2019). Mellander (2019) also noted that the number of employees in the Swedish gym industry skyrocketed in this period. She discovered large municipalities with big populations and high population density levels to have the most employees per capita within these establishments. This indicates that the gym activity levels are higher in the municipalities matching these characteristics.
12
2.4 Educational Attainment, Socioeconomic Status, and Health
The relationship between educational attainment, socioeconomic status, and health is well documented and the consistent findings suggest that highly paid, well-educated individuals have significantly healthier lifestyles than those of a lower education level and SES (Folkhälsomyndigheten, 2019b; Ross & Wu, 1995). Ross and Wu (1995) discovered that well-educated individuals, in general, were more physically active and had healthier consumption habits than individuals of a lower human capital stock, and some researchers even argued that health and lifestyle may be key components in determining the attained educational level of an individual in adulthood (Koivusilta et al., 2003).
The education-to-health relationship is also present in Sweden as the Swedes of a high human capital level are more likely to engage in physical activities and exercise during their leisure time in comparison to the Swedes of a lower educational attainment (Folkhälsomyndigheten, 2019a). Furthermore, the well-educated individuals in Sweden tend to have healthier nutrition habits as they are more frequent consumers of fruits and vegetables, and less likely to be regular smokers (Folkhälsomyndigheten, 2019b; Livsmedelsverket 2016). Although the health of the Swedish population in general is improving, the differences across regions are increasing. These differences are correlated with differences in education and socioeconomic status as the municipalities characterized by a high SES also tend to be the municipalities with the highest fraction of active individuals. The major difference in lifestyles between high human capital individuals and the less educated is mirrored in other health-related factors such as average life expectancy, obesity, and incidences of myocardial infarction in which the well-educated Swedes show considerably healthier values (Folkhälsomyndigheten, 2019b).
Florida (2002c) argued that the awareness of staying fit and living healthy among highly educated individuals was due to aesthetic reasons. He claimed that these individuals exercised to obtain the perfect bodies in order to appear more presentable to employers and potential partners. Yigitcanlar et al. (2007) also emphasized physical appearance as a key factor behind the healthy lifestyle habits of the high human capital individuals. They suggested that knowledge workers are tempted by environments in which they get to see
13
others’ perfect bodies meanwhile also getting the opportunity to display their own. Whether aesthetics is considered a driver of physical activities in a Swedish context or not is yet to be established.
3. Hypothesis and Expected Results
The hypothesis is formulated to investigate whether there is a correlation between the growth of health amenities (represented by the percentage change of employees within these establishments) and the share of high human capital individuals in the Swedish municipalities. Based on the literature review, I expect to find a positive correlation between the growth of health amenities and the fraction of high human capital individuals in the Swedish municipalities. A summary of the expected signs of each variable is presented in table 1.
Table 1: Variables, definitions, and expected signs
Variable Definition Expected sign % of
Employees in Health Amenities
Median Income The median income level of the daytime
population within a municipality after taxes +
Municipality Categories A categorical dummy variable showing whether the municipality is a dense municipality, large city municipality, or rural municipality
+
Population Density The number of residents per km2 in a
municipality
+
Share of High Human Capital Individuals in the Population
Share of municipality population (ages 25-64) with three years or more of higher education
+
Share of Working Population in Creative Industries
The fraction of the working population (ages 16-64), working in industries classified as creative by the Swedish Agency for Economic and Regional Growth
14
High human capital individuals have been discovered to be more likely to have healthier lifestyle habits than those of a lower education as they amongst other things tend to eat healthier and exercise more than less educated individuals (Folkhälsomyndigheten, 2019b; Ford et al., 1991; Ross & Wu, 1995). Furthermore, these individuals have been found to fuel the growth of amenities in the cities they reside as they tend to have a higher preference for amenities and higher affluent incomes to spend on these amenities in comparison to less educated individuals (Brueckner et al., 1999; Moretti, 2003; Shapiro, 2006). Hence, I expect municipalities with a higher presence of high human capital individuals to see a higher positive percentage change of employees in health-oriented amenities.
H1: “Municipalities with a higher fraction of high human capital individuals will
experience a higher positive percentage change of employees in health amenities”.
This is the main hypothesis tested in this thesis. Additional beliefs of mine are that large city municipalities, municipalities of higher socioeconomic status, and municipalities with high population density levels also will experience a higher positive percentage change of employees within health-oriented amenities.
15
4. Methodology
This paper aims to investigate whether there is a relationship between the fraction of high human capital individuals and the growth of health amenities across the Swedish municipalities. The data used to examine the hypothesis of this paper is retrieved from Statistics Sweden and the Swedish Standard Industrial Classification (SNI) 2007 register. I will use the data to analyse the percentage change of employees in health amenities in all 290 Swedish municipalities between the years 2010-2019.
4.1 Empirical Model
To conduct my analysis, I will use an empirical model describing the linear relationship between the dependent and independent variables. The model consists of cross-sectional data and the dependent variables are based on the percentage difference in the value of the given variable between the years 2010-2019. The independent variables are solely based on values from the year 2010. This model ensures that there is no problem with reverse causality since a percentage change of employees in the different kinds of health amenities over the years cannot affect the value of the independent variables. Based on the literature review, the following model is developed:
% lnEmployees in Health Amenities(2010-2019) = 0 + 1 (lnMedian Income) + 2
(Municipality Category)Di + 3 (lnPopulation Density) + 4 (lnShare of High Human Capital Individuals) + 5 (lnShare of Working Population in Creative Industries) +
I decided to use an ordinary least square regression as the estimation method for my model. To decrease the non-normality in my model, I used the natural logarithm (ln) on all my variables except for the categorical dummy variable. To control for multicollinearity between my explanatory variables, the Pearson correlation coefficients (Appendix 1) and the variance inflation factors (Appendix 2) were reviewed. The results indicated some issues with multicollinearity between the different measures of human capital. Hence, I decided to run these variables in separate regressions against a so-called “base equation” consisting of the other independent variables. This is thoroughly explained in section 5. Furthermore, the correlation matrix indicated that some of the other explanatory variables were somewhat highly correlated, but as the VIF values did
16
not signal for any severe multicollinearity problems, I decided to run these variables in the same equations. To inspect for heteroskedasticity, I used the Breusch-Pagan/Cook-Weisberg test. The results (Appendix 3) indicated that the regression residuals did suffer from heteroskedasticity in all equations. To overcome this problem, I ran the regressions with robust standard errors which helps remove the heteroskedasticity through a consistent covariance matrix estimator (White, 1980).
4.2 Variables
Dependent Variables
The dependent variables in my thesis intend to measure whether the growth of health amenities is related to the share of high human capital individuals in the Swedish municipalities between the years 2010-2019. The term health amenities refers to a combination of various amenities connected to a healthy lifestyle and is based on the number of establishments of these amenities within each municipality. Although many researchers previously have examined the links between environmental characteristics and the accessibility to amenities, goods, and services linked with a good health (Carr et al., 2011; Kawakami et al., 2011), I am not aware of any prior research using the term health amenities as a joint term for these goods and services.
While choosing which establishments to be considered as health amenities, I selected those which I naturally connected to a healthy lifestyle (fitness facilities, sports clubs, health food stores, etc.). Subsequently, I added certain establishments that have been used in previous research or are of a similar nature to establishments that have been used in prior research by Kawakami et al. (2011) that did not occur to me initially such as dispensing chemists. Furthermore, I chose not to include ski facilities as a health amenity in this paper as I believe that the growth of these facilities is contingent on peculiar areal characteristics that are only available in a very limited number of municipalities. The establishments denoted as health amenities in this thesis and their SNI-codes are listed in appendix 4. I also divided the health amenity measure into two categories based on whether the establishment is linked with physical exercise or retail of health-related goods and services to examine if the growth of the different types of health amenities were differently related to the fraction of high human capital individuals. The health amenities connected with physical activities and exercise are referred to as sport amenities whereas
17
the health amenities linked with the retail of sporting goods, fruits and vegetables, and other goods and linked to a healthy lifestyle are denoted retail-oriented health amenities The health amenities are sorted by category in appendix 5.
Due to the lack of accessible data on health-related amenities (i.e. number of tennis courts, number of swimming pools, etc.), the growth of the health amenities available was a relatively infrequent event in certain Swedish municipalities within this period of time. This resulted in low correlations with the independent variables when running the regressions with the percentage change in the number of health amenities as the dependent variable. Hence, I decided to use the percentage change of employees within these amenities rather than the percentage change of establishments to measure the growth of health amenities. This also applies when measuring the growth of sport amenities and retail-oriented amenities.
The data for my dependent variables is collected from Statistics Sweden and includes all of the 290 Swedish municipalities. In table 2, the municipalities who experienced the highest positive and negative percentage changes of employees in health amenities between 2010-2019 are presented. The characteristics of the municipalities with the highest positive percentage changes differ rather much as four of the municipalities are large city municipalities and dense municipalities whereas two out of the top ten municipalities are rural. However, all the ten municipalities with the highest percentage drop of employees in health amenities are rural. These municipalities are characterized by low population density, amenity levels, and poor socioeconomic status and have in recent decades struggled to attract and retain high human capital individuals (Bjerke & Mellander, 2017). The municipalities with the highest positive and negative percentage changes of employees in sport amenities and retail-oriented health amenities are available in appendix 6 and 7.
18
Table 2: Municipalities with the highest positive and negative percentage changes of employees in health amenities
Municipalities (Highest) % of Employees in Health Amenities Municipalities (Lowest) % of Employees in Health Amenities Tanum 175.30% Ragunda -79.37% Mullsjö 165.41% Laxå -78.08% Lidingö 140.68% Munkfors -68.63% Habo 91.53% Arjeplog -64.32% Kungsör 63.09% Kinda -59.14% Danderyd 60.80% Dorotea -57.92% Hammarö 55.84% Högsby -56.52% Huddinge 55.57% Bjurholm -56.49% Järfälla 55.41% Överkalix -54.20 Färgelanda 54.09% Norsjö -54.00% N=290
Source: Statistics Sweden 2019
Independent Variables
I have selected the independent variables for this paper based on what previous research has found to be positively correlated with the growth of amenities and a good health. The share of high human capital individuals and the share of the working population in creative industries are both measures of human capital levels. Individuals of a high human capital stock have been positively associated with both a healthy lifestyle and the growth of amenities (Folkhälsomyndigheten, 2019b; Shapiro, 2006).
High socioeconomic status is also associated with a good health profile and the growth of amenities (Folkhälsomyndigheten, 2019b; Ford et al., 1991 Moretti, 2003). Thus, the median income is included as a control variable. Large city municipalities generally have big population sizes that allow for specializations that are not feasible in smaller populations. Moreover, these municipalities are known to attract a large share of the high human capital individuals who as previously mentioned are known to boost the growth of amenities. Meanwhile, rural municipalities are struggling to attract and retain talented individuals (Mellander & Bjerke, 2017). To compare how the difference in the characteristics of the different municipality types affect the growth of health amenities, a categorical dummy variable is included. Finally, population density is included as an independent variable as dense areas have been found to experience a more rapid population growth which should also increase the demand for amenities (Glaeser et al., 2001).
19
Median Income
This explanatory variable measures the median income level of the daytime population after taxes of individuals in the Swedish municipalities. The individuals included in the measure are between the ages of 20-65 years old. Income level and socioeconomic status are widely known to be linked with educational attainment, healthy lifestyle habits, and a high valuation of amenities (Folkhälsomyndigheten, 2019b; Ford et al., 1991; Moretti, 2003). The variable is expressed in 1000s SEK and is based on data from Statistics Sweden.
Municipality Categories
This variable is a categorical dummy variable dividing the Swedish municipalities into three different categories developed by Growth Analysis (Tillväxtanalys, 2014) on behalf of the Swedish government. The three categories are rural municipalities, dense municipalities, and large city municipalities. The definitions of these municipality types are available in appendix 8 and the municipalities are sorted by category in appendix 9, 10, and 11. Large city municipalities are typically associated with dense, well-educated populations with a high preference of health whereas rural and dense municipalities tend to struggle to maintain and attract well-educated individuals, and have more problems with obesity and physical inactivity (Folkhälsomyndigheten, 2019b). This variable is meant to compare how the different municipality types and their characteristics are related to the growth of health amenities.
Population Density
The population density variable measures the number of inhabitants per square kilometer in the municipality. This variable is included as areas of a high population density have been found to experience a faster population growth than less dense areas, and as the population grows so does the demand for amenities (Glaeser et al., 2001). Furthermore, high population density levels might attract high-technology industries with well-educated employees who have been found to spur the growth of amenities (Kotkin, 2001; Shapiro, 2006). This variable is based on data retrieved from Statistics Sweden.
20
Share of high human capital individuals in the population
This variable measures the share of a municipality’s population (ages 25-64) with three years or more of higher education. This is the conventional way of measuring human capital (Mellander & Florida, 2007). The variable is calculated by dividing the number of individuals aged 25-64 with a bachelor’s degree or higher in a municipality by the total number of individuals within this age span in the given municipality. These individuals have been found to spur the growth of amenities and to have a preference for a healthy lifestyle (Shapiro, 2006). This variable is also referred to as human capital level and the conventional measure of human capital in this paper. The data for this variable was retrieved from Statistics Sweden.
Share of Working Population Working in Creative Industries
This variable represents the share of the working population (individuals between the ages 16-64) working in creative industries. This is a human capital measure used by Florida (2002c) who argued that it is more important to measure what people do than their educational attainment. Florida refers to this human capital measure as creative capital and he denoted the holders of this capital the creative class. This measure is calculated by dividing the number of individuals aged 16-64 working in the creative industries, categorized by the Swedish Agency for Economic and Regional Growth (Tillväxtverket, 2018), by the total working population in the municipality. This variable is based on information from various SNI 2007 industrial codes published by Statistics Sweden 2007. The industrial codes included in this variable are presented in appendix 4. This variable is also denoted as the creative class measure of human capital in this thesis.
21
5. Empirical Results
5.1 Descriptive Statistics
Between 2010-2019, the number of employees within health-oriented amenities increased in 139 out of the 290 Swedish municipalities and decreased in the rest. The average percentage change measured 0.49%, and the percentage change differed tremendously across the Swedish municipalities. Tanum experienced the highest positive increase of 175.30% whereas Ragunda saw the most significant decrease as the number of employees working in health amenities dropped by 79.37% in the municipality during this time period.
The number of employees in sport amenities increased in 154 of the Swedish municipalities, measured the same in 4 municipalities, and decreased in the remaining 132 municipalities between 2010-2019. Norberg municipality experienced an infinite percentage increase as it increased from 0 employees to 1.8 employees in sport amenities over this period of time. Thus, Norberg was excluded from regressions 3 and 4. Not taking Norberg into account, Mullsjö experienced the highest percentage increase of 269.86%. Laxå suffered the greatest percentage loss of employees in sport amenities as the number of employees dropped by 82.02%. The mean percentage change of employees in sport amenities measured 5.28% between 2010-2019.
The number of employees within retail-oriented health amenities increased in 127 municipalities and decreased in 160 municipalities. Tanum had the greatest increase of 309.29% whilst Ragunda suffered the greatest percentage decrease in employees measuring -85.23%. The number of employees in retail-oriented health amenities measured the same for the years 2010 and 2019 in Kil, Sölvesborg, and Älvsbyn.
The severe difference between the municipalities can to some extent be explained by variations in the explanatory variables. Descriptive statistics for the dependent and independent variables are listed in table 3.
22
Table 3: Descriptive Statistics
Variables Minimum Maximum Mean Std. Deviation
% Employees in Health Amenities -0.794 1.753 0.005 0.328 % Employees in Sport Amenities -0.820 2.699 0.053 0.390 % Employees in Retail-Oriented Health Amenities -0.852 3.093 -0.013 0.483 Median Income 181.70 310.80 210.79 16.41 Dense Municipalities 0 1 0.452 0.498
Large City Municipalities 0 1 0.100 0.301
Population Density 0.2 4504.30 136.85 473.55
Rural Municipalities 0 1 0.448 0.499
Share of High Human Capital Individuals
0.081 0.491 0.069 0.005
Share of Working Population in Creative Industries
0.004 0.050 0.015 0.07
N= 290, 289
Source: Statistics Sweden 2019
5.2 Correlation Analysis
When inspecting the bivariate correlation (Appendix 1) between the independent variables, I discovered a high correlation between the two measures of human capital, and previous researchers have pointed out that the creative class is highly correlated to high human capital individuals (Glaeser, 2005; Hansen, 2007). Hence, I decided to run the dependent variables against these measures of human capital in separate equations along with a base equation. This base equation consists of all the independent variables besides the measures of human capital, that being median income, population density, large city municipalities (dummy), and rural municipalities (dummy). There were also somewhat high correlations between high human capital individuals, population density, median income, and large city municipalities. However, as the values of the variance inflation factors (Appendix 2) did not indicate any severe presence of multicollinearity, I decided to run these variables together in the equations. In a previous version of this paper, the median house price was included as an independent variable to further measure socioeconomic status. However, due to high correlations, especially with the human capital level and median income, the variable was removed from the regressions.
23
5.3 Regression Analysis
To investigate to what extent the independent variables are related to the percentage change of employees in health amenities, a series of OLS regressions are performed. To examine whether sport amenities and retail-oriented health amenities are related differently to the explanatory variables, the health amenity variable is divided into these two categories. These categories are then run as the dependent variables in separate regressions. Besides from the categorical dummy variables, all variables are logged by their natural logarithm (ln) to decrease the non-normality. To avoid multicollinearity, the two measures of human capital are run in separate regressions. Dense municipalities function as the reference variables for the municipality categorical dummy variables in the regressions. The F-statistics indicate that all regressions are significant at a 1% level. The regression results are provided in table 4.
24
Table 4: Regression results
Variables Eq(1) Eq(2) Eq(3) Eq(4) Eq(5) Eq(6)
Constant -0.290 (0.584) -0.614 (0.584) -0.396 (1.550) -1.448 (1.512) 0.671 (2.411) -0.910 (2.458) Large City Municipality (dummy) -0.094*** (0.024) -0.090*** (0.024) -0.225*** (0.063) -0.210*** (0.065) -0.192** (0.088) -0.169* (0.089) Median Income 0.073 (0.105) 0.156 (0.104) 0.130 (0.278) 0.360 (0.271) -0.052 (0.444) 0.292 (0.443) Rural Municipality (dummy) -0.023 (0.017) -0.045*** (0.017) -0.056 (0.046) -0.107** (0.048) -0.025 (0.063) -0.103* 0.060) Population Density 0.035*** (0.007) 0.036*** (0.007) 0.072*** (0.022) 0.076*** (0.021) 0.086*** (0.020) 0.091*** (0.021) Share of High Human Capital Individuals 0.115*** (0.032) - 0.269*** (0.092) - 0.406*** (0.098) - Share of Working Population in Creative Industries - 0.074*** (0.020) - 0.150*** (0.057) - 0.225*** (0.084) N 290 290 289 289 290 290 R2 0.327 0.309 0.230 0.220 0.232 0.218 F-Statistics 22.92*** 26.94*** 15.78*** 14.89*** 16.45*** 13.47***
*** Significant at 1% level **Significant at 5% level * Significant at 10% level Robust standard errors within brackets
In equations 1 and 2, the percentage change of employees in health amenities is the dependent variable. Equation 1 examines the human capital level against the base equation. The results indicate that all explanatory variables besides median income and rural municipalities are significant at a 1% level. Population density, large city municipalities, and the measure of human capital are particularly significant throughout the equations which is reflected in the t-statistics (Appendix 12). According to the coefficient values, the share of high human capital individuals in the population is the
25
strongest explanatory variable for variations in the dependent variable whereas population density is the weakest. The R2 value of equation 1 measured 0.327 which is the highest R2 measure of all equations.
In equation 2, the share of the working population in creative industries replaces the conventional measure of human capital. In this equation, all explanatory variables except median income are significant at s 1% level. The coefficient value for the share of the working population in the creative class measures 0.074, which is notably lower than the coefficient value of the other measure of human capital in equation 1. The R2 value of the given equation measures 0.309. The large city municipalities dummy variable has a coefficient value of -0.090 in equation 2 which is the highest for all explanatory variables in this equation. Population density has the lowest coefficient of the significant variables in the equation.
In equations 3 and 4, the dependent variable is the percentage change of employees in sport amenities. As with equation 1, equation 3 runs the share of high human capital individuals with the base equation. The results illustrate that large city municipalities, population density, and the share of high human capital individuals have a significant relationship with the dependent variable at a 1% level whereas the remaining variables are insignificant at a 10% significance level or higher. The coefficient value of the human capital measure is more than twice as high as in equation 1, and a one percent increase in human capital level in equation 3 would imply a 0.269% increase of the dependent variable. As in equation 1, the measure of human capital and population density are the variables with the highest and lowest coefficient values respectively. This implies that the share of high human capital individuals is the main determinant of variations in the dependent variable whereas population density is the weakest determinant out of the significant variables. The R2 value of the equation measures 0.230 which is notably lower than previous R2 Values.
Equation 4 runs the creative class measure of human capital with the base equation. In this equation, all independent variables besides median income are significantly related to the dependent variable at a 5% level or higher. The dummy variable for rural municipalities is significant at a 5% level in the equation whereas the other explanatory
26
variables are significant at a 1% significance level. The coefficient value of the share of the working population in creative industries measures 0.150, meaning that a one percent increase in this variable would have a positive effect on the dependent variable by 0.150%. The coefficient values illustrate that the variables with the highest and lowest coefficient values remain the same as in equation 2. Even though the coefficient value is higher for rural municipalities than for population density, it is difficult to tell which of these is the stronger determinant since the population density variable has a more significant relationship with the dependent variable. The R2 value of equation 4 measures 0.220.
In equations 5 and 6, the percentage change of employees in retail-oriented health amenities is the dependent variable. Like equations 1 and 3, equation 5 tests the human capital level with the base equation. The results of this equation are rather similar to those of equations 1 and 3 as the same variables remain significant throughout these equations. Although, the large city municipality dummy drops to a 5% significance level in equation 5 whereas the other variables stay significant at a 1% level. The coefficient of the human capital variable measures 0.406 in this equation, which is the highest value in all equations for both measures of human capital. As in equations 1 and 3, the share of high human capital individuals in the population remains the strongest variable in explaining variations in the dependent variable. Large city municipalities rank as the variable with the second-highest coefficient value in this equation. However, the significance of this variable has dropped to a 5% level which implies that the accuracy of this measure is not as high as for the other variables. Population density has the lowest coefficient value out of the significant variables in this equation. The R2 value of equation 5 measures 0.232.
In equation 6, the share of the working population in creative industries is tested against the base equation. All the independent variables show a significant positive relationship with the dependent variable except for median income. The municipality dummy variables are both significant at a 10% level whereas the remaining explanatory variables are significant at a 1% level. The coefficient value for the share of the working population in creative industries measures 0.225 which is the highest for this measure throughout the equations. This is also the highest coefficient value for all significant variables in equation 6, indicating that the creative class measure of human capital is the main determinant of
27
changes in the dependent variable in this equation. The coefficients indicate that population density is the weakest determinant in this equation. Although, population density has a significant relationship with the dependent variable at a 1% level whereas the municipality dummy variables are only significant at a 10% level. The R2 of equation 6 measures 0.218 which is the lowest of all equations.
Although there are some differences between the regressions, the overall results do not change a lot between the different equations. Both the conventional measure and the creative class measure of human capital remain significant at a 1% level throughout all six equations. Furthermore, median income stays insignificant in all equations. The conventional human capital measure remains the variable with the strongest explanatory power for changes in the dependent variable in all equations it is included whereas the creative class measure of human capital is only the dominant variable in the last equation. The dummy variable for rural municipalities is the only variable shifting between being significant and insignificant throughout the equations. Although, the significance level drops for large city municipalities in the final two equations. The coefficients vary rather much across the equations and the values are more extreme in the later equations. This was expected as the sample is smaller for these equations. This also explains the big drop in R2 between equations 2 and 3. The coefficients are higher for the conventional measure of human capital than for the creative class measure for all equations. The results are more thoroughly discussed in the discussion and analysis section of this paper.
28
6. Discussion and Analysis
In this section of my thesis, the hypothesis is answered, and the regression results are discussed and connected to previous research, findings, and theories.
Median Income
The median income variable is part of the base equation and was thus included in all six regressions. This variable did not have a significant relationship at a 10% or higher with the dependent variable in any of the six equations. I found this result rather surprising as numerous studies have found individuals of a high socioeconomic status to have healthier lifestyles than those of a lower SES (Folkhälsomyndigheten, 2019b; Ford et al., 1991). However, inspecting the correlation matrix (Appendix 1) I noted that median income is rather highly correlated to population density and the share of high human capital individuals. This suggests that this variable might be insignificant since the income effects more or less already have been accounted for by the other independent variables. To examine if this was the case, I ran the regressions without the measures of human capital. The results indicated that median income had a significant, positive relationship with the dependent variables at a 5% level in the first two regressions, suggesting that some of the positive effects of median income had been captured by the human capital variables. However, the median income variable remained insignificant in the case when retail-oriented health amenities was the dependent variable. The regression results are available in appendix 13.
A further possible explanation for this could be that median income is measured on the daytime population (the population present in an area during business hours, including employees) of the municipalities whereas the human capital level is measured on the nighttime population (the population that resides in the municipality at night) in the Swedish municipalities. This could somewhat skew the measurement as it is likely that high human capital individuals spur the growth of amenities in the municipalities where they live rather than where they work. Previous research has indicated that Swedes of a high human capital stock have increasingly been concentrating in the major cities of the country (Mellander, 2016). Lundberg (2006) discovered that positive net migrations to one municipality also benefits the net migration of adjacent municipalities through
29
spillover effects, and it is reasonable to think that high human capital individuals move to adjacent areas close to the larger cities and commute to work rather than moving to the cities themselves to avoid the urban wage premiums and obscure house prices in these areas (Bjerke & Mellander, 2017).
Municipality Category
This municipality category variable was included to examine whether the dependent variables had experienced different changes across the different municipality types. Dense municipalities were used as the reference variable for the equations. The results varied for the different municipality types throughout the equations. Rural municipalities were negatively correlated to the dependent variable in all equations but were only significant in the regressions in which the creative class measure of human capital was used rather than the conventional measure of human capital. This negative correlation was rather anticipated since previous research has found rural areas to struggle with physical inactivity, obesity, and maintaining and attracting individuals of high human capital levels who highly value amenities and healthy lifestyles (Folkhälsomyndigheten, 2019b; Mellander & Bjerke, 2017). Despite that rural municipalities overall were negatively connected with the dependent variables, Färgelanda and Tanum were amongst the ten municipalities which experienced the highest positive percentage changes of employees within health amenities between 2010-2019. The positive changes of these municipalities were more or less entirely driven by an increase of employees within retail-oriented health amenities as Färgelanda experienced an increase of 153.85% and Tanum saw a 309.29% increase of employees within these establishments.
The results for the large city municipalities dummy variable, on the other hand, were not expected. Large city municipalities are positively correlated to all dependent variables (see Appendix 14), but the regressions tell us otherwise. The regressions results indicate that large city municipalities are significantly negatively correlated with the percentage change of employees in health amenities in all regressions. This does not align with the findings of previous research as large city municipalities previously have been connected to a high share of well-educated individuals, high socioeconomic status, and population density which are all components recognized to drive the growth of amenities (Jacobs, 1961, 1969; Mellander, 2016).
30
As is the case with median income, the large city municipalities variable is rather highly correlated with the share of high human capital individuals and population density. Although the VIF values (Appendix 2) did not signal for any severe multicollinearity, this variable appears to suffer from the positive effects it is commonly associated with being captured by the population density and the share of high human capital individuals variables. To investigate whether this is the case, I ran the dummy variables in separate regressions with the dependent variables. The regression results (Appendix 15) showed that large city municipalities are positively correlated with the dependent variable in all equations at a 5% significance level or higher, indicating that some of the positive effects have been assigned to the other variables in the regression. Furthermore, the mean percentage change of employees within health amenities in large city municipalities measured 22.9% whereas the average measured 8.1% in dense municipalities and -12.2% in rural municipalities. This suggests that large city municipalities are positively correlated to the dependent variables but that the positive effects were canceled out by other explanatory variables. A further explanation for the negative value could be the higher initial number of employees within health amenities in these municipalities in comparison to other municipality types.
Population Density
The population density variable is part of the base equation and was thus included in all regressions of this thesis. In equation 1, population density is run against the base equation with the percentage change of employees in health amenities as the dependent variable. The results suggest that population density is significantly positively related to the percentage change of employees in health amenities at a 1% level, and the coefficient indicates that for each 1% increase in population density, the dependent variable sees a positive change of 0.035%.
Population density remains positive and significant at a 1% level throughout all six equations. The regression results show that the population density variable is stronger for the regressions in which retail-oriented health amenities act as the dependent variable, suggesting that the percentage of employees within retail-oriented amenities increases more for each percentage increase in population density in comparison to the employees in sport amenities. Furthermore, the results show that the coefficients of the population
