Why are some Cities Winners or Losers?

Table of Contents

Table of Figures

1. Introduction

2. Theories of City Growth and City Size Distributions
2.1 Literature Review
2.2 Which Theory fits Germany?

3. Determinants of City Growth
3.1 Basic Determinants
3.2 Demographic Factors
3.3 Social Factors
3.4 Geographic Factors
3.5 Infrastructural Factors
3.6 Institutional Factors
3.7 Economic Factors
3.8 External Factors
3.9 Basic Factors

4. Uniting the Findings into an Overarching Model

5. Testing the Model
5.1 Included Data
5.2 Tested Variables
5.3 Excluding Outliers from the Model
5.4 Results for the First Regressions
5.5 Multiple Regressions
5.5.1 Model 1
5.5.2 Model 2
5.6 Optimizing the Model
5.6.1 Model 3
5.6.2 Model 4
5.7 Further Modifications

6. Interpretation of the Results
6.1 First Main Finding
6.2 Second Main Finding
6.3 Third Main Finding
6.4 Fourth Main Finding

7. Ideas for Further Research

8. Conclusion

Literature

Attachment 1: Overarching Model of City Growth

Attachment 2: Variables

Attachment 3: Analysed Data

Attachment 4: City Size Distributions, Migration and City Growth Over Time

Attachment 5: Simple Regressions Including Outliers

Attachment 6: Simple Regressions Without Outliers

Attachment 7: Multiple Regressions

Attachment 8: Migration for Different City Sizes

Table of Figures

Figure 1: Simple representation of the overarching city model

Figure 2: Linear regression of city size and average rents, Berlin included

Figure 3: Quadratic regression of change in real available income (2001-2011) and relative net migration

Figure 4: Cubic regression of nearby population growth and relative net migration

Figure 5: Linear regression of diversity and internationality

1. Introduction

Within only a few decades, Detroit lost over 60% of its inhabitants. It is a classic example for a city that has lost a big share of its populace, with mostly the unemployed and poor left after the exodus. Large areas of the city are now uninhabited. Unfortunately, the Detroit phenomenon is not unique: The same, albeit at a lower velocity, is also happening in some German cities, such as Duisburg. Here, whole neighbourhoods are being torn down, because most houses there are empty. The city government does not hope to increase its number of inhabitants anymore, or keep it constant: the “Duisburg 2027” strategy only aims to slow down the shrinking. Only a few kilometres away is the booming city of Düsseldorf located, with ever rising population levels and expensive rents. But this small distance is enough, and Duisburg’s decline is still unstopped.

In hindsight, many journalists, politicians, and economists appeared to be able to explain why all this had happened, but would it not have been better if they had been able to predict the emigration before it happened, so that counter-measure could have been taken?

The aim of this bachelor thesis is to find out what drives city growth in Germany. It shall find out what are the determinants that make large cities gain or lose inhabitants.

Especially in Eastern Germany after the reunification there have been some “winner cities” such as Dresden, Leipzig, and Potsdam, and many “loser cities” such as Chemnitz, Magdeburg, and Schwerin. In the last decades, a similar separation between winners and losers has arisen also in Western Germany. So what is the reason for this division into two groups of cities?

The research is very relevant for city governments and real estate owners or investors. Emigration of the working populace and few births lower the tax earnings and raise the necessary public expenses a city has to bear, while the opposite makes a city thrive fiscally, financially, and economically. Rents and land prices depend on the development of the city size: shrinking cities have horrendous vacancy rates, while prices are exploding in growing cities. Knowing what the reasons for city growth or shrinkage are enables city governors to adapt the right policies to make their city become an exception of the overall demographic collapse in Germany or to turn around a dangerous progress. Knowing these reasons also enables real estate investors to predict future demand for residential, office, and retail space, and to decide whether properties are under or overpriced.

The structure of this thesis is as follows: First, existing theories of city growth are compared in Chapter 2, and it is found out which of them fits Germany best. In Chapter 3, the determinants of city growth that have been identified in the existing literature are listed and explained. In Chapter 4, the findings of the literature are merged into an overarching model, which shows all the relations between the variables and city growth. Chapter 5 includes the statistical analysis of the most important variables for all German cities that have more than 250,000 inhabitants (or used to have at some point in time after 1990). In Chapter 6, the results found in Chapter 5 are condensed and interpreted. In Chapter 7, ideas for further research are presented, followed by a short conclusion in Chapter 8. The overarching model created in this thesis (in a complete and a simplified form), a table of the suspected determinants of city growth, and the inputs and outputs of the statistical analysis of German cities are included in the attachments.

2. Theories of City Growth and City Size Distributions

2.1 Literature Review

There are different theories on how cities grow differently and why there are characteristic distributions of city sizes among many countries. I have condensed these theories to three schools of thought. They differ in what they regard as the underlying reason for city growth, what they view as the fundamental reason for differences in growth, if they accept the idea of an optimal city size, and what they think is how city sizes react to external shocks, and how city sizes must be distributed. Bosker, Brakman, Garretsen, and Schramm (2006) and Redding and Sturm (2005), among others, give good overviews of competing city growth theories.

The first school of thought is what I would like to call the „returns to scale theory“. It says that the determinants of city growth are endogenous. For example, productivity could depend on population level, and a high productivity in turn could spur population growth, so that there are increasing returns to scale. This school of thought foremost views differences in economic potential as the reason for different population growth among cities. For the „New Economic Geography“, founded by Krugman (1991), market access (i.e. access to consumers and suppliers) is the main factor, as it leads to economic growth and thus population growth (and this in turn enlarges the market access). Connected with this school of thought is the belief that optimal city sizes exist (i.e. a population level where the marginal disadvantages of the city size become so big that they equal the marginal advantages), and that there are multiple equilibria. Because of that, when an external shock hits the city (such as a catastrophe), the population level can change so much that a different equilibrium is reached. If the number of inhabitants is only changed slightly, and the city will not pass the „threshold“ to a different optimal city size, it will return to its old equilibrium and therefore old population level. It could be compared to the atom model of Niels Bohr, where electrons can only be located in distinct orbits, and a certain amount of the electron's potential energy must change for it to move to another orbit. This means that an external shock can (but does not need to) lead to a different stable city size. Without external shocks, city growth is characterized by increasing or decreasing returns to scale. Therefore, when looking at the distribution of city sizes in a country, we are expecting to see divergent growth (i.e. larger cities grow faster or shrink slower, for example due to more innovation and human capital connected with the knowledge spillover in large cities) or convergent growth (i.e. smaller cities grow faster or shrink slower, for example due to high levels of congestion and pollution associated with city size, which reduce the living quality in large cities). Gonzales-Val and Olmo (2011) observed self-reinforcing increases for both income and city size in the USA after a certain income threshold, and they also found evidence for two equilibria at which there is an optimal city size (pp. 19-20).

The second school of thought is the „random growth theory“. It assumes that there is no fundamental reason for city growth and also no optimal city size to which the population level could converge. The difference in population growth among cities derives predominantly from institutional factors (e.g. clear zoning laws or policies aimed to attract immigrants^[1] from other cities). Because the growth is random, and an optimal city size does not exist, there is no „correction“ after an external shock. Shocks have a permanent effect on the city size. They do not change the overall shape of the city size distribution, though, because also shocks are assumed to be hitting random cities. Therefore, only the ranks of individual cities change. Apart from external shocks, the cities grow parallel. Gibrat's law, which states that the growth is independent of the initial size, is in effect. As long as there are no external shocks, usually Zipf's law holds for the city size distribution (at least for the large cities). This law states a distribution in which the largest city is twice as big as the second-largest, thrice as large as the third-largest, and so on. In practice, Zipf’s law was confirmed in many studies. Sharma (2003) deemed this school of thought to offer a correct depiction of city growth in India: He observed that differences in growth rates are only short term and caused by exogenous shocks, while in the long term cities grow parallel, with their growth being the same as the country’s overall growth rate (p. 319). Cordoba (2008) noted that many different countries in the world have Pareto distributed city sizes, and that it is necessary for this to have a balanced growth path and a Pareto distribution for the underlying source of growth (p. 177). Eaton and Eckstein (1994) also noted parallel city growth in France and Japan (p. ii), as did Eeckhout (2003) for the USA (p. 1448). Gabaix (1999) found out that shocks have a greater impact on small cities (pp. 760-761). Nitsch (2005) did a meta-analysis of 29 studies and concludes that the “estimated exponent of a Zipf regression is on average not 1.0” (p. 97), which means that cities are more evenly distributed than suggested by Zipf’s law.

The third school of thought is the „locational fundamentals theory“. The underlying reason for city size and city growth is the city’s location and its first nature geography, which leads to a productivity that is given and fixed. Differences in city sizes can be explained by different natural endowments and circumstances, such as climate, the existence of a navigable river, and the occurrence of diseases. Because everything derives from (mostly) unchangeable natural circumstances, there can be only one optimal city size. Analogously, external shocks only have a temporary effect and sooner or later the city returns to its old size. Furthermore, no rules for the overall distribution of city sizes can be inferred. In accordance with this school of thought, Cronon (1991) argues that the geography and location were the main determinants for Chicago’s development.

2.2 Which Theory fits Germany?

Brakman, Garretsen, and Schramm (2004) assessed the effect of the bombings of German cities in the second world war, and found out that the effect on city growth was only temporary in West Germany, but not in East Germany (which was attributed to the establishment of the GDR, in which free movement of people was not possible) (p. 202). They concluded that „The relative size of individual cities and the resulting city size distribution are remarkably stable over time for most countries. Evidence for some cities suggests that even wars or other large shocks do not always change the relative size of cities significantly over time“ (p. 201). This would corroborate the locational fundamentals theory.

Nearly the same authors, Bosker, Brakman, Garretsen, and Schramm (2006), which assessed German city growth from 1925 until 1999, found out that Gibrat's law only held for 25% of the cities, even when the changes in population were corrected for the Second World War (pp. 4-5). This implies that the random growth theory is wrong for Germany, because of the rejection of the independence of initial city size and population growth in 75% of the cases. It should be noted that again cities of the former GDR were excluded because of lacking availability of data and the fact that „firms and workers were not free to move between cities“ (p. 5). Because of the permanent effect the war had on more than 70% of the sample cities, the locational fundamentals theory is also not supported (p. 25). After 1945, smaller cities grew faster, because large cities were more destructed. Interestingly, before the Second World War the growth was independent of city size, which fits the random growth theory. After the war, there was a converging city size distribution, as only the largest cities grew independently of city size, while the (less war-torn) smaller cities grew faster (p. 18). The authors concluded that city growth in Germany is best described with decreasing returns to scale for the postbellum period (p. 27).

In a similar paper, Bosker, Brakman, Garretsen, and Schramm (2007) later took spatial interdependencies between cities into account when researching the effect of the war on city sizes. Their main finding was that the West German cities only recaptured half of the war shock by 1963. There was a tendency to move back to the pre-war rank, but not completely (p. 162). Therefore, the authors assumed that German cities are best described as having two equilibria (p. 165). Another finding of the article is that Munich was the economic centre of West Germany (e. g. cities close to Munich grew faster, cities close to Hamburg grew slower) (p. 161). In total, the returns to scale theory fit best to their observations.

The returns to scale theory is also supported when looking at the division of Germany. Redding and Sturm (2005) found out that West German cities near the border lost inhabitants (or grew slower) due to the reduced market access, and they were able to reject all alternative explanations. The subsidies for the „Zonenrandgebiet“ (the area adjacent to the GDR) were not able to stop the development, which shows the limited effect of institutional efforts. In contrast to Bosker et al.'s (2006) finding, Redding and Sturm (2005) observed that cities with more war destruction grew faster until the 1960s.

Together, these research results suggest that the city growth in Germany is best described by the returns to scale theory. That means that city sizes are neither fixed nor independent of themselves. City growth has endogenous causes and is affecting itself.

3. Determinants of City Growth

Cities are created by man. This implies that the reasons why people move into a city, or how many children they get, are too complex to fully understand. We can only hope to find the most important variables and approximate the many relationships between them. Still that does not mean that the models we try to create are correct. Henderson and Wang (2005) furthermore state that the effect that determinants may have on the growth of individual cities are very heterogeneous (p. 25).

When measuring the variables we propose, it is important to keep in mind that what we can determine are only correlations. This can have five basic interpretations: A causes B, B causes A, A and B mutually reinforce each other, another variable C causes both A and B, or the correlation is purely coincidental.

Therefore, one must not forget that everything this thesis, or any other, can achieve is bringing some structure into a reality that is too complex to grasp.

3.1 Basic Determinants

The variable of interest is city size (in terms of inhabitants), which can also be expressed as the change over time, namely population growth. There are three things that can change the city size: net migration, natural growth, and historical events.

Net migration is the difference of immigration and emigration (keep in mind that this first and foremost refers to inter-city migration, not only inter-country migration). It is the most important factor for German cities today, as the differences in fertility among cities are small and all cities have fertility rates clearly below what is needed to hold the population size constant without immigration (Teuber & Wedemeier, 2013, p. 10).

The difference of fertility and mortality (usually expressed as births and deaths per 1000 inhabitants) is what I call the natural growth. According to Goldman (2011), fertility is mainly determined by three factors.

The first one is religiousness. Obviously, “be fruitful and multiply” is part of many religions. But even beyond that, religion enables people to become “immortal”: Not only through the DNA carried by their children, but also through direct promises of the religion (such as eternal life or rebirth), and the inter-generational communication of culture. When a culture (including religion) dies, its people do not want to have children anymore, because they sense that only the biological immortality is left. This is Goldman’s explanation for the fact that shortly after religion died in the Western civilization, also the birth rates plummeted.

The second factor is the economic value of children. In agricultural societies without social security provided by the state you need children to have them work on your fields and care for you when you are old. Children thus have an economic value. In a modern society with social security, you just have the costs connected with raising children, but you do not get any benefits. The taxes for social care can also be paid by other peoples’ children. Thus you would be smart to have no children and act as a free-rider. The economic value of children is negative in Germany, and even child benefits from the state cannot change this. (The only exception would be people completely on welfare, who get all of their money from the state and thus can increase their income by having more children.)

The third factor is the educational and employment opportunities for women. This should be obviously bad for fertility, as career and family are not easy to combine, and it can easily be seen by comparing the situations of women and fertility rates in different countries.

Another factor for both fertility and mortality is the age structure. The more young people and less old people there are, the higher is the fertility and the lower is the mortality.

There are many more things that influence fertility, such as the availability of contraception, the average age of marrying, optimism or pessimism concerning the future, etc. Because of the complicated nature, and the overall low fertility rates and bad age structures in Germany, the determinants of natural growth are not empirically analysed in this work.

3.2 Demographic Factors

Religiousness was already mentioned; the effect it has on fertility can easily be seen on a map, as the birth rates in the more atheist East Germany are significantly lower than in the West.

The “youngness” of the age structure has a profound effect on the natural growth rate, too. On the other hand, it may also lead to more crime (Ellis, Beaver, & Wright, 2009), which in turn could make a city less attractive for immigration.

According to Capello (2001), who researched city growth in Italy, population density has an influence on quality of living (p. 25). The more attractive the living quality is, the more it fosters immigration. Density might lead to more criminality (Harries, 2006; Sampson & Groves, 1989), but there is no consensus in literature (Ellis et al., 2009).

Diversity of people is good for German city growth, as it spurs more migration from abroad (Teuber & Wedemeier, 2013, p. 22). Regional development is fostered by the presence of highly qualified foreigners and cultural diversity (Bellini, Ottaviano, Pinelli, & Prarolo, 2008, p. 34; Teuber & Wedemeier, 2013, p. 22; Niebuhr, 2006^[2], p. 13; Saxenian, 2006, p. 99). On the other hand, it also leads to more criminality (Ellis et al., 2009).

The population growth of a city leads to an expectation for future population growth. There is inertia in city growth, which implies that it is possible to infer future growth rates from past growth rates (Hernando, Hernando, & Plastino, 2013, p. 5). The expected population growth can have an effect on housing prices (DiPasquale & Wheaton, 1996, p. 57), which in turn can influence the decision to move to a city or not.

Hernando et al. (2013, p. 5) also found out that the growth rate of a city can be predicted from the growth rate of neighbouring cities. The interaction is normally limited to a distance of 70 km. This may reflect the fact that close cities merge into a larger metropolitan area over which the overall net migration can spread out.

3.3 Social Factors

The quality of living is one of the most important social factors and is determined by agglomeration economies. It is made up of positive and negative aspects of living in a certain city. Positive effects, which are called “city effects” by Capello (2001, p. 20) and hereafter, include, among others, better energy efficiency and better public services infrastructure (e. g. more school per km²). Negative effects, also called “urban overload”, include, among others, pollution, waste, and congestion (p. 20). A high quality of living makes a city attractive for people to move there (Rappaport, 2004, pp. 577-578; Xu & Zhu, 2008, p. 22), while a low quality of living will lead to a decrease in immigration rates (Sharma, 2003, p. 318). Basic factors for quality of living are population density and city size, but there are many more.

Criminality is another determinant for city attractiveness, as high criminality lowers the quality of living and thus makes it less attractive for immigration. If this is an important factor for changes in city size is questionable for Germany, but, for example, criminality significantly hinders city growth in Brazil, according to da Mata, Deichmann, Henderson, Lall, and Wang (2006, p. 27). Criminality is increased by diversity, unemployment, low incomes, bad education, a young age structure, a larger city size, and possibly a higher population density (Ellis et al., 2009).

Not only the quality of living makes a city attractive for immigration: also the relative quality of living in nearby cities needs to be taken into account. Whereas normally a city profits (in terms of a growing population) from other cities nearby (Hernando et al., 2013, p. 5), it can also happen that this effect is weakened or even turned negative, if the nearby cities have a much higher living quality. Why would you want to live in a polluted, dirty, criminal city, if there was a much more liveable city 10 km? Glaeser, Scheinkman, and Shleifer (1995, p. 5) observed this “voting with the feet” phenomenon in the USA.

Another factor that makes cities attractive in terms of living quality is culture. It is especially important for educated people (Teuber & Wedemeier, 2013, p. 21). Cultural attractivity can be separated into production of culture (measured for example through number of theatres, museums, and art schools) and, to depict the quality of the produced culture, consumption of culture (measured for example through the number of sold tickets for operas, museums etc.). Both components of cultural attractivity are fostered by a large city size, but also smaller cities can be culturally attractive (Teuber & Wedemeier, 2013, p. 21).

Besides these variables that measure the attractiveness of living in a city, education also plays an important role for population growth.

The educational infrastructure is usually measured as the number of universities in a city or (arising thereby) the share of students of the whole population. Obviously, having a university draws students into the city and thus increases the population level. Consequently, educational infrastructure has been an important promoter of city growth in the United States (Beeson, DeJong, & Troesken, 2001, p. 697). Furthermore, students often stay in the city after they receive their degree. This leads to a higher human capital in the city.

Highly qualified people are one of the most important factors for city growth, because firms need them as employees (Gonzales-Val & Olmo, 2011, p. 19). Education (or human capital) leads to a better access to the production factor “labour”, which in turn increases the attractiveness of a city for firms (Teuber & Wedemeier, 2013, p. 6). Furthermore, studies have shown that the presence of much human capital leads to more urban productivity in Brazil and the USA (da Mata et al., 2006, p. 4; Glaeser et al., 1995, p. 17; Glaeser & Saiz, 2003, p. 42), higher incomes and purchasing power in the USA (Glaeser et al., 1995, p. 1; Glaeser & Saiz, 2003, p. 42; Gonzales-Val & Olmo, 2011, p. 19), more innovation (Glaeser & Saiz, 2003, p. 43; Teuber & Wedemeier, 2013, p. 23), less unemployment (Glaeser et al., 1995, p. 18; Poelhekke, 2009, p. 20), and less criminality (Ellis et al., 2009). Consequently, human capital is associated with increasing city size in many countries, including the United States, Germany, and China (Beeson et al., 2001, p. 697; Glaeser & Saiz, 2003, p. 42; Poelhekke, 2009, p. 20; Wang & Zhu, 2013, p. 52). Au and Henderson (2005) were not able to find a relationship between education and city growth in China, but attribute this to the bad data they used (p. 571). Eaton and Eckstein (1994), who researched city growth in France and Japan, observed that migrants moving to a city are often less educated than the average, but they acquire human capital very quickly once they live in the city; emigrants are more educated than the average (pp. 34-35). Glaeser and Saiz (2003) note that skilled people are especially important in declining areas, because they enable the city to better adapt to the changing circumstances (p 43).

While education may increase net migration both directly and indirectly, the education level of women has a negative effect on fertility (Goldman, 2011). For example, it is common knowledge that female academics have below average fertility rates in Germany.

Also of importance is the role that a city plays in relation to other cities. In Germany, cities are divided by the federal states into three categories: Oberzentrum, Mittelzentrum, and Unterzentrum (upper, medium, and lower tier regional centres) . These terms refer to the degree in which a city has a central place status. This includes the availability of special goods and the presence of certain institutions such as hospitals, administration offices of the state, special warehouses, colleges, concert halls, football stadiums, etc. Institutions like these need high scale economies (for example, only in few cities a symphony orchestra makes sense, but petrol stations and supermarkets do not need a high degree of scale economies). The higher the central place status of a city, the higher is the quality of living there. Thus, Hsu (2012), who assessed city growth in China, suggests the importance of the central place status for a city’s attractiveness. The central place status is determined mostly by initial city size, which implies increasing returns to scale in this respect (p. 923).

Internationality of a city, measured for example by the number of nights’ stay of foreigners, the share of students that come from abroad, or the share of foreign workers, is a very important factor for German city growth, according to Teuber and Wedemeier (2013). Not only does it directly lead to more net migration (as, for example, students from abroad prefer to study in cities that are already very international), it also leads to a better access to potential employees for firms, which in turn implies an increased attractiveness of a city for companies. Furthermore, internationality leads to more network integration, and increases productivity in innovative industries (pp. 24-26).

Network integration refers to the degree in which a city’s inhabitants are connected to people from other cities. Capello (2001) measured network integration by the number and duration of phone calls to other cities and countries. A higher network integration gives firms access to a larger market (both consumer and supplier markets) (p. 23). Market access in turn increases the attractiveness a city has for firms, so that more of them will locate there. Therefore, it is not surprising that network integration brings about an increase in urban productivity (Teuber & Wedemeier, 2013, p. 22). Furthermore, Capello (2001) found a connection between network integration and a higher share of the tertiary economic sector in Italy. He also observed that higher network integration weakens the effect that a great number of economic activities increase urban overload (pp. 23-24).

3.4 Geographic Factors

The location of a city and its geography shape the development of its economy and demography: Gallup, Sachs, and Mellinger (1998) observed that countries far from a coast and in tropical regions are disadvantaged due to high transport costs for international trade and high disease burdens. This is negative for both the income level and income growth; yet those countries will have the highest population increases, although only due to fertility. Cronon (1992) assessed the history of Chicago and concluded that its location enabled the city to play the role as a gateway city to the hinterland. It was “the entrance and exit linking some large region with the rest of the world” (p. 307). Location and geography affect a city directly and indirectly (e. g. population growth and industrial composition).

Hernando et al. (2013) found out that the population growth rate of a city can be estimated from those of the cities within a 70 km distance (p. 1). Therefore, the closeness of other cities needs to be taken into account. While Au and Henderson (2005) found no spillover effects with neighbouring cities in China (pp. 570-571), they have been observed in both Europe and the USA (Bottazi & Peri, 2003; Rosenthal & Strange, 2004). The number of neighbouring cities has implications for the phenomenon of “voting with your feet”. Cheshire & Magrini (2002) regard both the relative economic attractiveness of and the relative quality of living in nearby cities as important variables that influence the immigration and emigration that a city experiences (p. 20).

Reachability is defined by Teuber and Wedemeier (2013) as the time needed to reach European agglomerations by car and plane. This variable is important for integration into a network of people living in different cities and the overall productivity. Reachability is mostly a question of infrastructure (for example, Berlin has a much higher reachability than Dresden, even though both have a similar distance to most European agglomerations) (pp. 26-27).

Access to natural transportation networks (such as oceans or navigable rivers) can also be seen as a factor for reachability, and thus lead to quicker city growth (Beeson et al., 2001, p. 697; Black & Henderson, 2003, p. 367). This may not play a role for people travelling, but for the transport of goods. Au and Henderson (2005) found the presence of navigable rivers to be insignificant for city growth in China (p. 571).

Climate was found to be a determinant of city growth in the early 19th century USA (Beeson et al., 2001, p. 697; Black & Henderson, 2003, p. 367). Personally, I do not think that it matters much for migration within Germany, as it is roughly the same in all regions, although I could imagine that in principle it has an influence on the quality of living.

Endowment with natural resources is another locational fundamental. It can be seen as a part of the access to production factors a firm has at a specific location (apart from labour).

3.5 Infrastructural Factors

Intercity infrastructure is another determinant of city growth. Da Mata et al. (2006) approximate the infrastructure’s utility with the intercity transportation costs and found an effect on population growth (p. 20). In the country of their research, Brazil, this could be improved by a great degree through bituminization of the highways. To quantify the intercity infrastructure’s utility, da Mata et al. (2006) measured the government’s investments in intercity road infrastructure (p. 10); Au and Henderson (2005), examining China, measured the investments in intercity rail infrastructure and the kilometres of paved road per capita, as well as the distances to major highways. They did not find a significant effect on population growth, though (p. 571). For America, studies show that the city size grows in correlation with the utility of the intercity infrastructure (Beeson et al., 2001, p. 687; da Mata et al., 2006, p. 20). Teuber and Wedemeier (2013) suggest that this is due to an increased reachability (pp. 26-27).

The second infrastructural factor is the utility of the urban infrastructure. Da Mata et al. (2006) see commuting costs and commuting time as expressions of this concept (p. 10). Xu and Zhu (2008), who examined Chinese city growth, measure this variable through paved road surface per capita (but this is probably not applicable to Germany, as there are so few urban unpaved roads). A good urban infrastructure leads to more income, either because of lower commuting costs or more time to work when commuting time is low (da Mata et al., 2006, p. 10). Quality of living is increased as well (Xu & Zhu, 2008), both via an increase in city effects (positive agglomeration economies) and a decrease in urban overload (most notably congestion) (p. 13). Eberts and McMillen (1999, p. 1491) and Xu and Zhu (2008, p. 13) add that a good public urban infrastructure increases firms’ productivity.

3.6 Institutional Factors

Rodrik, Subramanian, and Trebbi (2004) analysed the income levels of different countries in the world, and discovered that institutional circumstances are more important than geography or trade (p. 131). Therefore, it would be logical if institutions also affected the development of cities.

Leadership capacity refers to the autonomy in decision making a city has. A good measure (adapted from Cheshire and Magrini (2002, p. 16)) would be the relative size of a city to its federal state. The three city states (Berlin, Hamburg, Bremen) enjoy the most autonomy in decision making, while other cities are subjugated to the will of the federal state they belong to. For example, the dozens of large cities in North Rhine-Westphalia have to live with what the federal state legislates (and obviously, big cities like Cologne and Düsseldorf are more likely to get their way than smaller cities like Bergisch Gladbach and Bottrop). According to Cheshire and Magrini (2002), leadership capacity is associated with productivity (pp. 16-17).

Leadership capacity alone is not sufficient, though: also leadership quality is needed. Da Mata et al. (2006) found out that Brazilian cities with better administered local land use and zoning laws experience a stronger increase in city size (pp. 23-24). Wang & Zhu (2013) suggest that good city management decreases the “city disease” and gives Tokyo as an example (which is both the largest and the safest city of the world) (p. 52). It thus makes sense to assume a relation between leadership quality and quality of living. Acemoglu, Johnson, and Robinson (2001), although they did not research cities but countries, also note the important role that institutions play for the economic development: with better institutions, more resources are invested in physical and human capital, which finally leads to a high income (p. 1369). In a comparison of 49 countries, La Porta, Lopez-de-Silanes, Shleifer, and Vishny (1998) also found evidence for a link between the legal systems (e. g. common law vs. civil law) and the economic development (p. 1152); but this becomes less relevant when only cities of one country are compared.

3.7 Economic Factors

The concepts of market access and access to production factors were already mentioned; together they make a location attractive for firms. Redding and Venables (2004) examined countries (instead of cities) and discovered that remoteness from markets impedes the settlement of firms (p. 78); the same logic should be applicable to cities as well. In the “New Economic Geography” approach, started with Krugman (1991)^[3] and further developed by Fujita, Krugman, and Venables (1999), the main determinant of city growth is market access (also called market potential). Market access includes both access to consumers and access to suppliers. A typical method to measure market access is to calculate the sum of population level divided by distance for every other city (plus the own city’s population):

It is thus a measure of the proximity a city has to other cities, taking account also the sizes of these cities. A high market potential thus implies many consumers and many suppliers that can be reached from a city. Transport costs and scale economies have led to the geographical concentration of economic activities (Hason, 2005, p. 22).

Access to a large market (and therefore a high attractiveness for firms) has led to a higher number of economic activities in Italy and Brazil (Capello, 2001, p. 23; da Mata et al., 2006, pp. 5, 27) and more productivity in China (Au & Henderson, 2005, p. 569). Redding and Sturm (2005), who assessed the development of West German cities near the inner-German border, blame the loss of market access on the population loss they had to experience. Less access to consumers meant a decline in wages, less access to suppliers meant a rise of the consumer price index. In consequence, the economic situation of the people (especially employees) worsened. People moved into more Western cities, which had a greater market access and thus higher incomes and lower living costs. This was especially true for working people, as for example retired persons only had to bear the rising prices (pp. 38-39). Hanson (2004, p. 23) and Redding and Venables (2004, p. 78) also observed that wages and incomes are higher in proximity to large markets. Henderson and Wang (2005, p. 25) and (for the USA) Dobkins and Ioannides (2000) as well as Black and Henderson (2003, pp. 366-367) also found that market access leads to an increase of net migration and therefore city size. This is probably due to the effect that market access has on a city's economic attractiveness for people.

Economic attractivity consists of two components: income and cost of living. It is an important cause of migration between cities and thus city growth (Glaeser et al., 1995, p. 1). Other effects of higher income are lower criminality (Ellis et al., 2009) and higher housing prices (Capozza, Hendershott, Mack, & Mayer, 2002, p. 23), which in turn mean higher cost of living, though.

The economic attractiveness of a city is also influenced by how economically attractive it is to live in a nearby city or in the countryside. A city's economic attractiveness should only be seen relative to others. For example, da Mata et al. (2006) found out that cities experience more immigration when the rural income opportunities sink, even when the urban incomes and prices stay the same (pp. 19-20).

Competition, measureable for example as the number of firms relative to workers, positively affected urban productivity in Brazil (da Mata et al., 2006, p. 24).

A high productivity in turn leads to more immigration into a city, at least it was observed in the USA and China (Rappaport, 2004, pp. 577-578; Xu & Zhu, 2008, p. 25). It is typically measured as the real GDP per capita. Au and Henderson (2005) define it as the value added per worker.

Productivity is also fostered by foreign direct investment. Foreign direct investment can be measured per worker or as the share of the GDP it has. Au and Henderson (2005, p. 570) as well as Xu and Zhu (2008, pp. 24-25) found foreign direct investment to be an important driver of productivity in Chinese cities. Furthermore, it increased employment (Xu & Zhu, 2008, p. 24). Therefore, cities with much foreign direct investment experienced a high population growth (Xu & Zhu, 2008, pp. 29, 31).

Innovation is another factor. It can be approximated by measuring the share that knowledge-intensive industries or R&D (Research & Development) have in the economy. Innovative cities experience a higher productivity (Teuber & Wedemeier, 2013, p. 15; Hall, 2001, p. 188). The overall effect on net migration is unclear though: Teuber and Wedemeier (2013) see knowledge-intensive industries as becoming more and more important in Germany, while cities depending on old industries (e. g. the secondary sector) face problems, both economically and concerning population growth. By contrast, Cheshire and Magrini (2002), who examined twelve European countries including Germany, found that the share of R&D in the economy is negatively associated with population growth, although the statistical significance is not given for every model of city growth.

Industry structure as a whole has indirect effect on population growth; when industrial composition changes, also the city size may change (Black & Henderson, 2003, pp. 366-367). In the United States, both the primary and secondary sectors were associated with low incomes, compared to the tertiary sector (Beeson et al., 2001, pp. 687-688; Glaeser et al., 1995, p. 1). But that is not the only advantage of the tertiary sector: a high share of it makes it easily possible to give work to rural immigrants in China (Xu & Zhu, 2008, p. 31), and in Italy it was found that service-dominated cities experienced less urban overload that manufacturing-dominated cities of the same size (Capello, 2001). The city effects were first decreasing with the share of the tertiary sector, but then increasing for the better part (p. 21). Cities that have similar industrial structures also have similar population levels and educational attainments in the USA (Black & Henderson, 2003, p. 366).

The number of economic activities is increasing both city effects and urban overload (Capello, 2001, p. 23). Hence, the final effect on the quality of living is inconclusive.

In theory, it would make sense if cities’ population levels rise as long as their employment opportunities are growing, too (Sharma, 2003, p. 318). Employment, or the absence of unemployment, is associated with the average income in American cities (Glaeser et al., 1995, p. 1). Au and Henderson (2005) found that in Chinese cities there is an “∩” (inverted U) shaped relationship between income and employment, which shifts with industrial composition and market access (p. 569). This might be explained by the idea that a high supply of labour might either lead to unemployment (while the wages remain constant) or to rising absolute employment numbers, but with decreasing wages. Employment also leads to higher housing prices (Cheshire & Magrini, 2002, p. 20), which means high cost of living. Still, unemployment is obviously not good for city growth. This is also corroborated by the findings of Xu and Zhu (2008) in China (p. 24). A side benefit of employment is that it is connected with a reduced crime rate (Ellis et al., 2009).

Housing prices, also called urban rent, real estate prices, or land rents in literature, are having an effect on the cost of living for most people. The only people that profit from it are the landlords. Capozza et al. (2002), who researched the real estate market in the USA, thus do not conclude on the effect that housing prices have on the overall economic situation of the people. For Germany, with its high share of renters, I assume that high housing prices are an overall economic burden to the people. High housing prices (or, alternatively put, high rents) lead to more construction. This in turn lowers the real estate prices and rents again. How much is constructed does not only depend on the initial real estate prices, though, but also on the construction costs (p. 23).

The overall national economy also has indirect effects on the development of city sizes. The national economic performance (measured as GDP per capita) is connected in both directions with the productivity of cities. Capello (2001) showed that an improvement in the national economy also led to economic growth in the individual cities. The larger cities (especially Milan and Rome) profited most from the upturn (p. 18). Of course, the national economic performance is highly dependent on the performance of the cities, and larger cities carry more weight than small cities or the rural areas. The national output composition also has an effect on urban productivity: da Mata et al. (2006) observed that if there are some growth sectors in Brazil, the cities specialized in these industries are benefitting. Which industries, and implicitly which cities, are the “winners” is sheer luck. Cities with a better education of the labour force can adapt better to these changes, though (p. 5).

The last economic factor is the economic value that children have, as it is very important for the fertility rate (Goldman, 2011). I think that the economic value of children is mostly the same throughout German cities. Children may be more valuable in very rural areas, though.

3.8 External Factors

The rural population growth is of course an important factor for urban population growth as well, as a significant share of the children raised in the countryside will later move to the cities, because of education or employment. The more children are born in the countryside, the more immigration cities will experience. This is especially apparent for developing countries with a high rural population growth, such as Brazil and China (da Mata et al, 2006, p. 18; Xu & Zhu, 2008, p. 31).

A final exogenous factor is historical events. These can be good or bad: For example, the world war bombings were bad for German cities, and the expulsion of the Germans from the Eastern territories (as horrible as it was) led to explosive increases in the city sizes of the FRG and GDR. The division of Germany led to emigration from the West German cities near the border, and the reunification led to a (although smaller) countermovement. Historical events can have either a direct effect on the population size (e. g. a catastrophe or war leading to many dead), or it influences migration patterns or the natural growth rate. To analyse the regular determinants of city size, historical events needs to be identified and subtracted out.

3.9 Basic Factors

City size is not only at the end of all these relationships and causation chains; it does also influence other variables itself. According to Bettencourt, Lobo, Helbing, Kühnert, and West (2007), the pace of social life is increased in large cities (p. 7301). They also notice a surprising global “universality in human social dynamics, despite enormous variability in urban form” (p. 7305). Overall, large cities have “established” themselves in the world, and therefore, when they lose inhabitants for some reason, have a slower decrease in absolute or relative city size (Black & Henderson, 2003, p. 366).

Bettencourt (2013, p. 1438), Bettencourt et al. (2007, p. 7301), and Eaton and Eckstein (1994) state that in general the wages are higher in larger cities. Au and Henderson (2005) also say that for Chinese cities “real incomes per worker rise sharply with increases in city size from a low level” (p. 549). Of course a city with high wages also attracts more immigrants; this is an example for increasing returns of scale. Nevertheless, Au and Henderson (2005) observed that the average income first rises with city size and then slowly decline past a peak (p. 549). Notably, the whole relationship between the two is changes with the industrial structure of the city. A concentration on “lower” activities such as manufacturing means that the income reaches its peak already at a lower population level. Why exactly the city size influences the incomes is difficult to answer. One could imagine, for example, that large cities mean access to a large market, which makes it possible for companies to pay better wages; at some threshold though, the presence of so many people leads to an oversupply of labour, which results in declining wages. Gonzales-Val and Olmo (2011) noted that large cities in the USA experience immigration of many unskilled people, which might at some point lead to stagnating or decreasing income, and is a reason for the varying income levels among big cities (pp. 19-20). Due to the complex nature of social coherences, it is practically impossible to really understand the causal relationships; all that can be done is to find patterns and speculate about the reasons.

City size obviously has an influence on the quality of living. An increase of the number of inhabitants generally leads to both more city effects and more urban overload (Capello, 2001, pp. 20-22)^[4]. For example, the per capita use of energy and other resources is typically lower for large cities; also the public services infrastructure (e. g. banks, post offices etc.) is better. On the other hand, large cities experience more congestion, pollution, and so on. This can be influenced by many other factors (such as good leadership), but city size is one of the main determinants. Especially the increase in urban overload in large cities is a well-known phenomenon, which has been observed, among others, in the European Union, the United States, and India (Capello, 2001, p. 21; Cheshire & Magrini, 2002, p. 10; Glaeser et al., 1995, pp. 6-7; Glaeser & Saiz, 2003, p. 42; Gonzales-Val & Olmo, 2011, p. 20; Sharma, 2003, p. 318). The quality of living is further reduced by criminality, which is more present in large cities (Bettencourt et al., 2007, p. 7305; Ellis et al., 2009).

A larger population is also leading to an increased productivity (Bettencourt et al., 2007, p. 7305; Eberts & McMillen, 1999, p. 1491). According to Eberts and McMillen (1999), firms like to go where other firms already are, because at those locations there are agglomeration economies: businesses can share public goods and use them as production factors (e. g. a common pool of labour, knowledge and expertise, and the urban public infrastructure in general). Especially manufacturing companies are more productive in big cities. The authors thus conclude that “productivity of firms vary by city size and the level of investment in public infrastructure” (p. 1491).

Research on different countries has shown that city size is furthermore positively correlated with housing prices (Capozza et al., 2002, p. 2; DiPasquale & Wheaton, 1996, pp. 56-58; Eaton & Eckstein, 1994, p. 32), cultural production and consumption (Teuber & Wedemeier, 2013, pp. 21-22), the number of economic activities present (Hsu, 2012, p. 923), human capital (Eaton & Eckstein, 1994, pp. 33-34; Wang & Zhu, 2013, p. 52), the central place status (Hsu, 2012, p. 905), and of course firms’ access to production factors, consumers, and suppliers (Teuber & Wedemeier, 2013, pp. 6-7; Redding & Sturm, 2005, pp. 38-39). Innovation is also increased by city size (Bettencourt, 2013, p. 1438; Bettencourt et al., 2007, p. 7305), but is also needed to sustain population growth: “As population grows, major innovation cycles must be generated at a continually accelerating rate to sustain growth” (Bettencourt et al., 2007, p. 7301). This means that growth driven by innovation implies no limit to city size, and therefore also the absence of an optimal city size (p. 7306).

A positive net migration does not only lead to more inhabitants, it also increases housing prices (DiPasquale & Wheaton, 1996, p. 58) and the expected future population growth (Hernando et al., 2013). Redding and Sturm (2005) also observed that it leads to lower unemployment, as the economically active population is more mobile and willing to move to another city. People on welfare, on the other hand, tend to stay in their home town, which means that economically motivated emigration leaves the unemployed behind (pp. 30-31). Furthermore, total migration (the sum of the absolute numbers of immigration and emigration) leads to a younger age structure. This is because young people (18-30 years) tend to move into the city (from the countryside or the small suburbian cities around the main city), while people over 30 tend to move out of the city (to the suburbs and bedroom towns beyond the city border) (Teuber & Wedemeier, 2013, p. 11; Redding & Sturm, 2005, p. 31).

A higher fertility rate does not only mean an increasing city size, but obviously also a younger age structure. A younger age structure leads to a higher fertility rate in turn. For Germany, this especially means that the “old” cities face a demographic death spiral, while the “young cities” have a more positive and self-reinforcing natural growth rate.

4. Uniting the Findings into an Overarching Model

The literature review has brought to light many relevant variables. I have explained the variables that have direct and indirect effects on population growth and the effects that population growth has on these variables again. Of course, one must be aware that some of these findings are for other countries, such as Italy, the United States, China, and Brazil. It is unclear whether they are also true for Germany or not.

Nevertheless, I combined all these variables and their interdependencies into one model. This diagram shows the model in a simplified form:

Abbildung in dieser Leseprobe nicht enthalten

Figure 1: Simple representation of the overarching city model

You can see a complete graphic representation of the model in attachment 1.

The only direct influences on city size or population growth are net migration, natural growth, and historical events (which will be disregarded in the analysis, as history does not exactly repeat itself and therefore it is not possible to derive any certain information for future city size development from them). I will refer to city size, net and total migration, and natural growth as “basic variables”.

Migration and natural growth are influenced by a number of variables (I will call them “first order variables”), namely urban economic attractivity, quality of living, innovation, educational infrastructure, closeness to other cities, diversity, employment, urban productivity, rural population growth, population growth of nearby cities, youngness of age structure, religiousness, educational and employment opportunities for women, and the economic value of children. The first order variables can influence each other and are influenced by other variables, including the basic variables.

Some variables are not directly influencing the basic variables, but only have an effect on the first order variables (or each other). If they depend on other variables, I will refer to them as “second order variables”, and if they do not depend on any other variables (at least not in the scope of existing city growth research) I will call them “third order variables”. Sometimes they are virtually independent, for example closeness to other cities or access to a navigable river are given and fixed. Other variables, such as foreign direct investment or religiousness do depend on other circumstances, but their causes are not included in the model. This is because it is generally a never ending procedure to find the underlying reasons for them, and in turn the reasons for the reasons, and so on. It is like a child asking “and why is that?” again and again to every answer it gets. As I explained before, the world is too complex to really understand it, and therefore it makes no sense to try to make the model more and more detailed in the false hope it would finally be a correct depiction of reality. As with fractals, “zooming in” more will still leave us with the same complexity. Even the “virtually independent” third order variables are not truly independent: for example, the closeness to other cities can be changed by the founding of new cities (such as Brasília or Nánhuì New City) or the death of existing cities, and rivers can be made navigable or substituted by an artificial channel; the decisions to build a channel or found a new city are not independent of the development of cities. Intuitively, we assume that there exist some truly independent natural or social variables, which are not in any way changed by the avalanche of opinions and decisions that people form and make in reaction to a change in city size or one of its causes; this does not mean those variables’ values are completely unchangeable, but that there is no feedback at all from the city model. It is unknown which might be the truly independent variables, though.

Sometimes, variables have different components. Take for example the urban economic attractiveness: this is an abstract concept and cannot be measured directly. It is thus divided into two components: income and cost of living. These can be measured easily.

Some variables, especially abstract concepts, can only be measured through a proxy variable. For example, the consumption of culture could be measured as the number of sold tickets for museums and operas, human capital could be measured as the distribution of educational degrees in the population, and the quality of infrastructure could be measured as the amount of money that is invested in it.

A table listing all the concepts, components, measurements, and effects used in the model can be found in attachment 2.

5. Testing the Model

5.1 Included Data

For a table of the analysed data, please refer to Attachment 3.

Cities with over 100,000 inhabitants are officially called “Großstadt” (large city), due to the International Statistics Conference of 1887. At that time there were only 11 Großstädte in Germany, while today there are 80. If a certain exclusiveness for the Großstadt-status is desired, it makes sense to adjust the threshold. Therefore, data was researched for all German cities with over 250,000 inhabitants and those that had this many inhabitants in 1990 and then fell under the threshold^[5]. This amounts to 30 cities. For a more thorough analysis, it will be nice to include also smaller cities in the future, for example all settlements that legally are a city, but that was not possible in the scope of this bachelor thesis. All cities that have or have had over 100,000 inhabitants since 1949 in the West or 1990 in the East (in total about 90 cities) were included for the calculation of some variables, such as market potential, though.

The data for those thirty cities was collected for the year 1991, 2001, and 2011. Earlier years were originally planned to be included, but this plan was then disregarded, because the development of city growth was distorted due to historical events: During the Second World War, most large German cities were heavily destroyed (e. g. Berlin, Essen, and Dresden), while some others were hit less (e. g. Wiesbaden). After the war, 12 million Germans were expulsed from the former German territories in the East as well as from other countries in Central Europe. These people had to resettle in the cities of today’s Federal German Republic. Flensburg, for example, saw its population rise from 67,000 before the end of the war to 101,000 after the war and the expulsion (today: 83,000). The division of the remaining German area into a Soviet-controlled and three Western-controlled sectors was another shock, as millions of people moved from the Soviet zone to West Germany. The division itself had an immense negative effect on West German cities near the border, as Redding and Sturm (2005) showed. At the same time, East Germans were not free to move between cities within East Germany. Due to these institutional constraints it makes no sense to research city growth in the GDR. The same is true for Berlin, which was divided into a communist Eastern part and a Western part with an administrative special status. During the 1970s, there were many rearrangements of the administrative borders in West Germany. Often, many small cities were merged into a big one, or a big city “acquired” the surrounding smaller cities. For example, Bergisch Gladbach had about 50,000 inhabitants until it merged with some other small cities in 1975 and then suddenly boasted about 100,000 inhabitants. The reunification in 1989/1990 was the last historical event that shocked city growth in Germany. First of all, it removed the problems the West German cities close to the border faced during the 40 years before. For the East German cities, it resulted in a huge exodus, as millions of people moved to the West due to better employment and income opportunities. There was no area in the former GDR that was not experiencing the immense emigration. It was not until 2012 that the West-East migration was balanced for the first time.

Therefore, I believe that the years before the reunification of 1990 were distorted too much by these events. It makes no sense to compare these cities with the aim to find a general city growth model. It would be necessary to subtract out all the effects these historical circumstances had, but it is not possible to distinguish rigorously between changes in city sizes resulting from the historical events and “regular” city growth. Thus, 1991 is chosen as the first year of observation (even though there was still much migration from the East to the West).

The best data was available for the year 2011, as it was the year of the first and only census in the reunified German state. Therefore, data from 2011 is regarded as the most reliable data available. In 2001, there were censuses in the whole European Union with the exceptions of Sweden and Germany. Instead, Germany conducted a “census test”, in which it was tried to find reliable data with alternative methods. Because of the census test, and the temporal placement just in the middle of the twenty years from 1991 until 2011, 2001 was chosen as the third year for which data was included in the analysis^[6].

The resulting data was three-dimensional: different variables were measured for different cities at different times. For most analyses, the same cities in different years were analysed independent of each other. The underlying idea was that city growth should always be dependent on the same determinants, independent of the year. If a certain determinant would have been only valid in a certain year, it would not be possible to draw conclusions for the future from it. Attachment 4 shows that net migration and overall city growth differed between the decades. Interestingly, city sizes seem to stay the same. A possible explanation could be that the cities had an increasingly negative natural growth, which was offset by increasing net migration.

Natural growth was not analysed, as the fertility rates are below replacement in all of the cities. Thus the main driver for natural growth is the age structure. Migration has a positive effect on the age structure, and therefore the analysis of migration is already the largest part of the analysis of city growth as a whole.

5.2 Tested Variables

The effect of the first order variables on migration were tested. The first order variables were economic attractiveness, quality of living, innovation, educational infrastructure, employment, productivity, closeness to other cities, rural population growth, population growth in nearby cities, diversity, and internationality. For Germany, diversity and internationality have been identified by Teuber and Wedemeier (2013) as determinants of city growth. The ambiguous effect that closeness to other cities may have was found by Cheshire and Magrini (2002) for twelve European countries including Germany. Nevertheless, these variables were also included in my analysis, because of possible interactions with other variables. Rural population growth was excluded, as it was only identified as a driver of city growth in developing countries such as Brazil, and it was assumed that in Germany the rural population growth has been quite constant in the last two decades. Virtually all rural areas in Germany (that are not very close to a big city) have been experiencing the same problems of emigration and superannuation.

Net migration was measured in two ways: absolute net migration (immigrants minus emigrants for each year) and relative net migration (net migration per 10,000 inhabitants). The effects of the other variables were tested on both measurements of net migration.

Economic attractiveness was also measured in different ways. The first was the available nominal income per year, which was published by the Federal Bureau of Statistics (also called “Destatis”) for the years 2011 and 2000 (Destatis, 2014). Using the official inflation rates of these nine years, the nominal available income of 2000 was adjusted to 2011 (real available income per year). Another measurement was the purchasing power index: Germany as a whole is set to 100 and the observed cities range from 83% of the German average (Halle an der Saale) to 134.4% (Munich). The purchasing power index for all cities could only be attained for 2011 (MB Research, 2012). Besides available income, also the cost of living was interesting in order to assess the economic attractiveness. Data for this was taken from UNICUM Verlag GmbH & Co. KG (2014). The average monthly living costs for students were published. This means that only college cities were observed, which excludes Mönchengladbach. Furthermore, the living costs were not dated; therefore they were allocated to 2011, which should be approximately right. Gelsenkirchen was reported to have living costs of more than 995 € per month, which seems odd, as it is more than 100 € more than the next most expensive cities (Düsseldorf, Frankfurt, Hamburg, and Munich, which are all by far more expensive than the remaining cities).

Quality of living was measured as the share of recreational area of total city area. This seemed to be a good overall measure of living quality, more so than for example fine dust pollution (which is not very varied between the cities, and exists nowhere in disturbing amounts) or some city effect measures like schools/banks/hospitals/etc. per capita. It is also much less dependent on city size and density than most measurements mentioned in the literature, such as Capello (2001, p. 20). The data for 2011 and 2000 could be retrieved from Destatis (2014).

Innovation was measured as the number of patents per 100,000 workers. Unfortunately, it was only possible to find numbers for half of the cities and only for 2013. This may limit the meaningfulness of the analysis. As an alternative measurement, firms’ investments per worker were available for 2011 and 2000 from Destatis (2014). According to Felder, Harhoff, Licht, Nerlinger, and Stahl (1994), investment intensity is in a close relationship to innovation intensity (pp. 46-47). It might thus be adequate to use the investment intensity as a proxy for innovation.

Educational infrastructure was measured as the number of universities. Only real universities, i.e. those with the right to award doctorates and post-doctoral lecture qualifications, were included, not technical colleges (“Fachhochschulen”) or “universities of applied science”. In 2011, Berlin had the highest number of universities (8) and only Gelsenkirchen and Mönchengladbach had none.

Employment was measured as the opposite of the unemployment rate. For example, a city with an unemployment rate of 10% has an “employment rate” of 90%. In reality, this does not mean that 90% of the inhabitants are employed, as for example children, pensioners, students, and housewives are not (necessarily) counted as unemployed. The data for unemployment in 2011 and 2001 were available from Destatis (2014).

Productivity was measured as the nominal GDP per capita for every city. The numbers were available from Destatis (2014) for the years 2011 and 2000.

Closeness to other cities was limited to cities within a radius of 70 km, in accordance with Hernando et al. (2013, p. 1). Together with nearby population growth, it aimed to account for the “growth shadow effect” that Cheshire and Magrini (2002) observed (p. 20). This variable could be problematic, as it depends on the administrative borders (e. g. most suburbs of Berlin lie within the city borders since the Great Berlin Act of 1920). Two measures were derived: the number of cities within the 70 km distance that have more than 100,000 inhabitants or had at any time in the FGR^[7], and the number of people living in these cities at each point in time. In total, 88 cities and their population levels were included in the calculation of these variables.

Nearby population growth was also calculated, using the population growth of those of the other 87 cities that were within a 70 km radius of the observed city. The average annual growth from 1991 to 2001 was allocated to the other numbers for 2001, and the average annual from 2001 to 2011 was allocated to 2011.

Diversity was measured as the share of non-citizens of the total population. The data was available from Destatis (2014) for the years 2000 and 2011. The amount of ethnic non-Germans (which would be a more appropriate measurement for cultural and ethnic diversity) is higher, though, as “foreigners” are legally defined through citizenship and information on naturalized foreigners are not diligently documented. Increased naturalization might also explain the slightly decreasing share of non-citizens between 2000 and 2011 in many cities. Derived from the first measurement, the changes in the absolute numbers of non-citizens from 2000 to 2011 were also calculated for all cities.

The last of the first order variables, internationality, was measured as the share of jobholders who are foreigners. It could be calculated from official Destatis (2014) data for the share of foreigners, unemployment among foreigners, and overall unemployment, although here no information for 1991 or the early 1990s was available, either.

Additional variables that were included in the analysis were housing prices (measured as the average monthly rent per m² in 2011 for each city), the number of inhabitants for 1991, 2001, and 2011, and the market access figures. For the market potential, the exact coordinates of the 88 biggest cities in Germany (see above) were retrieved and used to calculate the distances between all these cities^[8]. The population numbers of all these cities were divided by the distance (the “own city’s” population was divided by 1) and the resulting fractions were then added. This was done for 1991, 2001, and 2011. In 1991, Augsburg had the lowest market potential; in 2001 it was Magdeburg and in 2011 it was Chemnitz (also the lowest overall). Berlin had the largest market potential in all three decades, although it has been decreasing. The follow-ups have been Hamburg, Munich, Cologne, and Essen in all years. This ranking also shows that the used market potential formula resulted in values very similar to the number of inhabitants.

5.3 Excluding Outliers from the Model

First, outliers were identified and removed from the analysis. (For the results of the regressions without exclusion of the outliers, please refer to Attachment 5).

Berlin’s city sizes and market potentials were excluded for all years. Its number of inhabitants was 783%, 769%, and 753% above the median value in 1991, 2001, and 2011, respectively. Furthermore, Berlin is nearly twice as big as the next biggest city, Hamburg. Because of that, Berlin was able to change the results of regressions dramatically:

Abbildung in dieser Leseprobe nicht enthalten

Figure 2: Linear regression of city size and average rents, Berlin included

[...]

---

^[1] In this thesis, the terms „immigrant“ and „emigrant“ refer to people moving to/from other cities, not necessarily other countries. Someone moving from Munich to Berlin would already be a migrant.

^[2] Niebuhr (2006) found that diversity is good for innovation, in the sense that it leads to better performance of regional Research & Development in Germany. A flaw of the study is that it only takes into account employed non-German immigrants, although there are many unemployed migrants.

^[3] Krugman (1991) wrote: “In order to realize scale economies while minimizing transport costs, manufacturing firms tend to locate in the region with larger demand, but the location of demand itself depends on the distribution of manufacturing. Emergence of a core-periphery pattern depends on transportation costs, economies of scale, and the share of manufacturing in national income” (p. 483). The preferred location of specific industries can be determined by other, specific factors, though (p. 498).

^[4] Capello (2001) furthermore found that in Italy the city effect growth declines after a city reaches 361,000 inhabitants, while urban overload is decreasing with city size for cities below 55,000 inhabitants (p. 21).

^[5] These cities are (decreasing in current population levels): Berlin, Hamburg, Munich (München), Cologne (Köln), Frankfurt am Main, Stuttgart, Düsseldorf, Dortmund, Essen, Bremen, Dresden, Leipzig, Hanover (Hannover), Nuremberg (Nürnberg), Duisburg, Bochum, Wuppertal, Bielefeld, Bonn, Münster, Karlsruhe, Mannheim, Augsburg, Wiesbaden, Gelsenkirchen, Mönchengladbach, Brunswick (Braunschweig), Chemnitz, Aachen, Halle an der Saale, and Magdeburg.

^[6] Sometimes, unavailable data for the year 2001 was substituted by data for 2000.

^[7] These cities are (besides the 30 cities of the overall analysis): Bergisch Gladbach, Bottrop, Bremerhaven, Cottbus, Darmstadt, Dessau-Roßlau, Erfurt, Erlangen, Flensburg, Freiburg im Breisgau, Fürth, Gera, Göttingen, Hagen, Hamm, Heidelberg, Heilbronn, Herne, Hildesheim, Ingolstadt, Jena, Kaiserslautern, Kassel, Kiel, Koblenz, Krefeld, Leverkusen, Lübeck, Ludwigshafen am Rhein, Mainz, Moers, Mülheim an der Ruhr, Neuss, Oberhausen, Offenbach am Main, Oldenburg (Old.), Osnabrück, Paderborn, Pforzheim, Potsdam, Recklinghausen, Regensburg, Remscheid, Reutlingen, Rostock, Saarbrücken, Salzgitter, Schwerin, Siegen, Solingen, Trier, Ulm, Wilhelmshaven, Witten, Wolfsburg, Würzburg, and Zwickau.

^[8] For the lines of longitude (which have the largest distance from each other at the equator and grow closer until they merge at the poles), a distance of 70 km for 1° was assumed. The resulting distances between the cities differ less than 1% from the real distances.