Agglomeration Economies

Get Complete Project Material File(s) Now! »

Empirical Findings

This section presents the empirical results of the study. It starts by providing a brief over-view of the data with descriptive statistics and a box plot. It is followed by a correlation analysis and ends with a discussion of the results from the two regression analysis.

Descriptive Statistics

The descriptive data presented in Table 3 gives an overview of the examined variables‟ minimum, maximum and mean values, together with their standard deviations. The varia-bles are all expressed in their original forms. The zero values in CultureFacilities, Culture2000 and Culture1990 are given a small value (0.001) when log transformed.
A box plot (see Figure 1) illustrates the municipalities with the highest concentration of cul-ture in relation to population. The line within the box represent the median, while above and below this line represent the upper and lower quartiles. The marked regions are outliers where Simrishamn (1.161) is the first one, just outside the maximum point.
Access to a high populated region is a common factor for the majority of the outliers. The municipality reaching the highest location quotient (2.751) of cultural workers is Stock-holm, which is followed by Lindingö and more within the Stockholm region, together with Sweden‟s third most populated municipality, Malmö.
Guthenburg is the second most populated municipality, but not noted as an outlier. It is situated right underneath the point to be marked as an outlier with a location quotient of 1.091. Älmhult, Simrishamn, and Falun are exceptions in terms of population size that all can be observed in Figure 1 as outliers despite their small population sizes.
The municipalities with the lowest concentration of cultural workers follow the same pat-tern. Municipalities with small population sizes have low location quotients. Perstorp, Bjurholm and Laxå are the bottom three. Bjurholm is Sweden‟s smallest municipality (2011) with 2431 inhabitants.
Culture2000 had similar outliers as Culture20113, Stockholm first, Lindingö second and Malmö third. Culture1990 had Gothenburg as the municipality with the highest concentra-tion of culture, followed by Lidingö and Malmö. However, the overall pattern is slightly different, in the sense that different municipalities are marked as outliers in 1990 compared to the top culture concentrated municipalities in 2011 and 20004.
Further by looking at Table 1 again, the top location quotient for Culture1990 and Cul-tureFacilities appears disproportionally high. Limited data with several municipalities having zero observations lowered the national average. Hence, the municipalities with the highest share of cultural workers in relation to their working population might have got a location quotient that is higher than what it would have got with more observations and overall bet-ter data. They are nevertheless not excluded in the models, since the relation between the municipalities still give indications of their influences on culture and cultural variety.
A frequency table of CulturVariety is found in the appendix5. Bjurholm and Åsele are the only two municipalities with registered workers in only one of the six culture occupations, while 19 municipalities have registered workers in all six occupations6.
The concentration of cultural facilities is highest in Botkyrka. Danderyd is the municipality with the highest average income, Årjäng the lowest. The local government in Helsingborg invests the largest share of their budget on culture (4.58 %), while Nordanstig only invest 0.96%, representing the national low. Haparanda got the highest regional share of foreign-borns (39.85%), and Lekeberg is the least diverse municipality in terms of foreign-born population (3.84%). Gothenburg got the highest grade on the Tolerance-index (4.23), while Tingsryd and Norsjö both got the lowest (1.00).

Correlation Analysis

The correlations between the dependent and independent variables, used in the model are summarized in a bivariate correlation matrix below. All variables are in their logarithm form.
All independent variables show a significant positive correlation to Culture. The variables with the highest correlation coefficients are POP (0.799), CulturalFacilities (0.661) and Cul-ture2000 (0.946). Furthermore, POP is highly correlated to the other independent variables, especially CulturalFacilities (0.714), Culture2000 (0.829) and Tolerance (0.831) thus, indicate multicollinearity.
Culture have the strongest correlation to Culture2000 (0.946), which indicates that culture have stayed in municipalities with already high concentration of cultural workers the past decade. Cultural workers appear to have been more mobile during the 1990s based on the much lower correlation between Culture2000 and Culture1990 (0.561).

Regression Analysis

This section is divided into two parts. The first section is a weighted least square regression the location pattern of cultural workers. The second section studies the variety of culture in an ordinal logit regression.

Weighted Least Squares Regression

Before the weighted least squares regression, an ordinary least squares regression including POP is performed. Culture2000 is the only variable showing a positive significant influence on Culture. However, except from introducing heterogeneity in the data, POP does not show a significant relationship to Culture.
Therefore, the second regression is a weighted least squares with POP as the weight varia-ble. All variables got a slight increase of their VIF values, but none reaches above 4. The R-squared increased and there are now three variables showing significant positive estimates to culture; Culture2000, CulturalFacilities, and AverInc.
The first hypothesis stated that cultural workers should be overrepresented in bigger cities with a high demand, based on the assumption that culture benefits from economies of scale. Average income shows a significant positive relationship to the location of cultural workers, which is in line with this and the previous literature. A higher average income suggests that people have more money to spend on culture and by that drives up a munici-pality‟s aggregate demand
On the contrary, population size did not show any significant relationship to the location of culture in 2011, which disagrees with previous research. The variable‟s strong correla-tion to Culture2000 (see Table 4), could suggest that cultural workers still, to a large extent, locate in densely populated regions, but that the location of culture in 2011 is more affect-ed by the location pattern of culture in the past decade, than of population size.
A significant relationship to culture in 2000 together with an insignificant relationship to culture in 1990 indicates that the location of culture today is affected by the location of cul-ture 10 years ago, but not as much by the location of culture 20 years ago. Moreover, popu-lation shows a higher correlation to culture in 2000 (0.829), than to culture in 1990 (0.663) in the correlation analysis (see Table 4). This might indicate that cultural workers‟ speed of urbanization were greater during the 1990s than between 2000 and 2011. However, this is not in line with the Swedish population as a whole, and nothing that has significant proof in this study. It is merely a discussion of the possible explanation regarding the population variable‟s insignificance in the regression model of this study.
The hypothesis that culture benefits from economies of scale and locate in more populated regions with a high demand cannot be rejected. Municipalities with an overrepresentation of cultural workers have a higher average income, which suggest that the hypothesis is not rejected, even though population size did not show a significant relationship to the concen-tration of culture.
The second hypothesis stated that municipalities with a local government who spend more on culture per capita will have a higher regional share of culture and cultural workers. GEXP is an average of local governments‟ expenditures on culture from 2007 to 2011. It does not show any significant relationship to cultural workers in 2011. Hence, this hypoth-esis is rejected. However, if the variable were to include more than just five years back in time, the result might be different.
The third hypothesis stated that the regional share of cultural facilities is expected to have a positive influence on the regional share of cultural workers. CulturalFacilties shows a signifi-cant positive relationship towards Culture, which means that increasing the share of local expenditure on culture, still is beneficial in order to attract cultural workers. However, this data does not say whether these facilities are financed by public or private investors. Hence, this thesis can only conclude that the local governments‟ expenditures on culture during the past five years have not affected the location of culture in a significant way. Even though cultural facilities show a positive relationship to culture, that is significantly more than zero.
If cultural facilities significance is due to scale or scope effects is difficult to say, but its high correlation to population (see Table 4) indicates that it is more likely to be a result of econ-omy of scale. Culture facilities represent live activities, which are getting more difficult to consume as distance increase. It is highly distance-sensitive and depends on local consump-tion.
The fourth hypothesis said that the place of culture in the past affects the place of culture today. It is assumed that past location patterns affect future location patterns. As been mentioned above, Culture2000 showed significant relationship to Culture, while Culture1990 did not. Culture2000 showed the strongest correlation to Culture (0.946) in the correlation analysis (see Table 4) and with a very high t -statistic reaching values over 20 (see Table 5), the place of culture in 2011 appears to be heavily dependent on the place of culture in 2000. The insignificant value of Culture1990 suggests that the path dependency of culture in 2011 were lower to 1990 than to 2000. The correlation between Culture and Culture2000 is considerably higher than the correlation between Culture2000 and Culture1990 (See Table 4).
The reasons why cultural workers were more path dependent the past decade compared to the 1990s, are difficult to determine based on this study. However, it is an indication of that path dependency of culture in Sweden has become stronger. The location patterns of cul-tural workers in 2011 are highly dependent on the location of cultural workers in 2000, while not showing a significant relationship to the location of cultural workers in 1990.
Nevertheless, this thesis can determine that culture has gotten more path dependent during the past decade, compared with two decades back. If this is a pattern that stretches further back in time cannot be determined in this study. It could be a result of the digitalized socie-ty reducing the importance of changing location, or the result of an increasing specializa-tion pattern in Swedish municipalities. The significant relationship to cultural facilities indi-cates that the second alternative is more likely. It is based on the idea that on overrepresen-tation of cultural facilities is a sign of specialization.
Control variables Diversity and Tolerance do not show a significant influence to Culture, de-spite showing significant correlation (See Table 4). Their insignificance support previous research on the “creative class” influence on creative occupations location patterns in Swe-den (Hansen and Niedmoysl, 2009; Asheim.) The reasons behind their insignificance are difficult to determine in this study.
Based on the significant values from CulturalFacilities, AverInc and Culture2000 more proba-ble reasons behind cultural clustering appear to be specialization and scale benefits, which supports Hallencreautz and Lindequvist (2002).

READ  Economic and financial determinants of persistent high-growth performance 

Ordinal Logit Regression

This sub section studies what variables that affect the variety of culture occupations. The dependent variable CultureVariety is tested against the same variables as in the WLS in order to investigate the hypotheses of economies of scope.
The “test of parallel lines” shows that the null hypothesis cannot be rejected, thus the pro-portional odds assumption is met and the ordinal logit model fits7. The threshold parame-ters represent the intercepts for each of the ordinal outcome. The intercepts can differ as long as the slope for each variable stays constant in a pattern of parallel lines (O’Connell, 2006). The odds-ratio is the exponential value of the estimated logits and represents the change in odds for a unit increase in one of the predictors. A value of 1 or close to 1 im-plies a non-significant relationship to the variety of culture.
The fifth hypothesis stated that the number of different cultural occupations expected to be positively related to population size and average income. It is assumed that scope effects are present when combined with economies of scale.
Population size and average incomes show positive significant values to the variety of cul-ture occupations, which mean that municipalities with a larger population size are more likely to have a broader range of cultural workers.
Since population size and average income represent economies of scale, this suggests that CultureVariety‟s benefits from economies of scope are likely dominated by benefits from economies of scale. Bigger regions with a higher average income and by that, a higher ag-gregate demand, are more likely to have a broader variety of culture.
The sixth hypothesis stated the number of different cultural occupations to be positively influenced by cultural facilities, based on the idea that different culture occupations wants to cooperate and coproduce in order to get around the venues‟ high fixed costs.
Cultural facilities do not show a significant result to the variety of culture. This is a further an indication of benefits of economies of scope to be a result of the benefits from scale ef-fects. Cultural facilities insignificant value to CultureVariety in comparison to its significant relationship to Culture (see Table 5) indicates that cultural facilities are driven by scale ef-fects more than scope effects
The seventh and last hypothesis stated that the variety of culture should be positively relat-ed to the place of culture in the past, based on the assumption of the location of culture to be a function of its past. The results support this by showing a positive significant value to Culture2000. The variety of culture appears to be dependent of the place of culture in 2000, but not on the place of culture in 1990, which is similar to the results from Table 5.

Conclusion

Based upon current research stressing creativity‟s increased importance for regional growth, this thesis aims to understand what economic factors that affect the place of culture and creative people in Sweden.
Research on why creativity and culture tend to cluster contains classical economic geo-graphic theories such as, economies of scale and economies of scope. Cultural workers or industries do according to theory benefit from scale effects, and should be drawn to re-gions with a higher aggregate demand. Cultural workers want to exploit economies of scope by cooperate with other cultural occupations in order to solve problems of expensive venues and receive mutual gain of knowledge spillover. Furthermore, research suggests that past specialization patterns, affect future specialization pattern. Regions with a history of culture are assumed to be more likely to keep their cultural heritage and have a higher re-gional share of culture today. A higher regional share of cultural facilities and local gov-ernments investing a larger share of their budget on culture are other factors mentioned to be positively related to culture in previous research.
The spatial distribution of culture is based on where in Sweden cultural workers lived in 2011. Two regressions are run, where the country is divided into its 290 municipalities. The first studies the concentration of culture in a weighted least squares regression. The results show that average income is positively related to location of cultural workers. The people in the municipalities with a higher average income have more money to spend on culture, thus increasing the demand. This is according to the literature and an additional evidence of that culture and creativity benefits from exploiting economies of scale. Moreover, the share of government expenditures spent on culture was not found significant to attract cul-tural workers. On the other hand, cultural facilities showed a positive relationship to the place of culture, which implies that municipalities with more money invested in culture, could still be a way to attract cultural workers. Furthermore, the place of culture in 2000 is also showing a positive relationship to the place of culture in 2011. This path dependency is an additional indication of specialization as a factor of importance, to attract cultural workers.
The second regression is an ordinal logit regression and investigates the variety of cultural occupations among the Swedish municipalities. The results showed significant positive rela-tionships of population, average income and culture in 2000 to the variety of culture. The variety of culture appears to be driven by similar forces as the concentration of culture. The significance of average income and population size together with a non-significant relation-ship to cultural facilities indicates that possible scope effects are more likely to evolve when combined with economies of scale.
To sum up, the main findings of this paper is that the economic geography of culture in Sweden is driven by scale effects, access to cultural facilities and past dependency. Howev-er, this paper could not find support on increased local government expenditures on cul-ture as a way to attract cultural workers. This should be investigated into more detail, using data that stretches further back in time. There is a probability that many of these cultural facilities are financed by the state, which should be looked further into and compared to those financed by private investors. A closer study of the historical pattern of path depend-ency in culture could also be useful. This paper found indications of cultural path depend-ency in Sweden to be weaker in 1990-2000 compared to 2000-2011. This possible pattern and underlying reasons should be investigated into more detail.

Table of Contents
1 Introduction
1.1 Delimitations
2 Literature Review
2.1 Agglomeration Economies
2.2 Why are Face-to-Face Interactions still important?
2.3 Path Dependency
2.4 Hypotheses
3 Method
3.1 Description of Variables
3.2 Methodology
3.3 Empirical Model
4 Empirical Findings
4.1 Descriptive Statistics
4.2 Correlation Analysis
4.3 Regression Analysis
5 Conclusion
References
GET THE COMPLETE PROJECT

Related Posts