Abstract
Popularised by the work of Richard Florida, the role of tolerance, openness and social or cultural diversity in urban development has gained much attention. Recent literature on urban and regional economics has found associations between these social factors and technology, entrepreneurship, innovation, housing and economic performance. In most of these studies, the terms tolerance, openness and diversity are generally conflated or interchangeably used. This article argues that diversity’s impacts on innovation and entrepreneurship are notably different from tolerance and openness and that diversity should be defined and measured differently from tolerance and openness. This article uses data of US metropolitan areas to examine the statistical difference between diversity and tolerance, and compares the effect of each on innovation and entrepreneurship in multivariate analysis. Diversity is measured using the Herfindahl–Hirschman index based on countries of birth, while tolerance is measured using the composite gay and bohemian index.
1. Introduction
Popularised by Richard Florida’s The Rise of the Creative Class (2002), the role of tolerance, openness and cultural or social diversity in regional economic development has gained much scholarly attention. Florida argues that tolerance signals low entry barriers; it therefore contributes to talent attraction of regions or cities. Since his seminal book, the effects of tolerance, openness and diversity on regional development have been explored from various perspectives. In addition to its association with attracting talent, tolerance has been connected to technology and innovation (Florida and Gates, 2001; Florida et al., 2008; Niebuhr, 2010; Qian et al., 2012), entrepreneurship (Audretsch et al., 2010; Cheng and Li, 2012; Lee et al., 2004; Qian et al., 2012; Smallbone et al., 2010), housing prices (Florida and Mellander, 2010; Ottaviano and Peri, 2006) and to economic performance that is typically reflected in wage, income, employment or gross regional product (Florida et al., 2012; Florida et al., 2008; Lee, 2011; Mellander and Florida, 2011; Qian, 2010).
Despite the growing recognition of the role of these social factors in regional and urban development, the terms diversity, tolerance and openness are generally conflated or interchangeably used. Florida (2003, p. 10) defines tolerance as “openness, inclusiveness, and diversity to all ethnicities, races, and walks of life”. He has primarily used three measures of tolerance in various studies—namely, the Gay Index, the Bohemian Index and the Melting Pot Index. These indexes are based on regional densities of gays and lesbians, bohemians and immigrants respectively. Page (2007, p. 332) contends that Florida’s Creativity Index, which combines his ‘3Ts’ (technology, talent and tolerance), “is at best a crude proxy for diverse toolboxes”. Similarly, Reese and Sands (2008) observe that although Florida uses different measures of population diversity to represent tolerance, the actual connection between diversity and tolerance is not well established (Reese and Sands, 2008, p. 5).
To address the criticism from Page (2007), Reese and Sands (2008) and others, this article compares diversity and tolerance. It argues that tolerance should be differentiated from diversity when we study the social or cultural underpinnings of regional and urban economic development. Tolerance and diversity are conceptually different and should be measured differently in empirical analysis. At the core of this study is the effort to reach a better understanding of the social drivers of innovation and entrepreneurship by comparing the effects of tolerance and diversity on innovative and entrepreneurial activities in US cities. We focus on innovation and entrepreneurship because they have been increasingly recognised as core sources of regional competitiveness (Acs, 2002; Fritsch, 2008; Malecki, 1994; Porter, 1998; Sweeney, 1987).
The article is organised as follows. The second section discusses the association between tolerance/openness/diversity and innovation and entrepreneurship. The third section distinguishes tolerance and openness from diversity in both concepts and measures. The fourth section compares the effects of tolerance and diversity on innovation and entrepreneurship in US cities using multivariate analysis. The fifth section discusses the empirical results and their implications on urban public policy. The last section proposes future research.
2. Innovation and Entrepreneurship: The Role of Tolerance, Openness and Social or Cultural Diversity
In this section, tolerance, openness and social or cultural diversity are interchangeably used, consistent with the literature and especially with Florida’s work. After tolerance and diversity are conceptually distinguished from each other in the following section, this theoretical part will be re-examined and the unique contributions of tolerance and diversity will be identified. There are various definitions and perspectives of innovation and entrepreneurship in the literature. It is worth clarifying that innovation involves not only the creation of new knowledge but more importantly its commercialisation. Entrepreneurship in this research follows the definition by Qian et al.
the discovery of market opportunities and appropriation of their associated market values via creating new firms (Qian et al., 2012, p. 2).
2.1 Tolerance and Innovation
Recent literature has identified three mechanisms by which tolerance may contribute to innovation. First, and along the line of Florida’s original work (2002), tolerance is associated with a region’s low barriers to entry, which encourages talent (i.e. human capital and the creative class) to move into a region or helps a region to retain talent. Tolerance is relevant in this context because of the background diversity among talent in cultures, countries of birth, religious beliefs, sexual orientations and so on. In addition, ethnically diverse cities feature diversities in cultures, foods and entertainment that may attract people who value diverse experiences (Olfert and Partridge, 2011). While human capital or the creative class drives innovation (Florida, 2002; Florida et al., 2008; Lucas, 1988; Romer, 1990), tolerance and diversity contribute to regional innovation indirectly via talent attraction and retention. The positive association between tolerance and talent has been found in a large body of empirical research (Boschma and Fritsch, 2009; Florida, 2002; Florida et al., 2012; Florida et al., 2008; Mellander and Florida, 2011; Qian, 2010).
Secondly, regional diversity indicates diverse available knowledge and diverse perspectives of thinking that are critical to innovation (Florida, 2002; Jacobs, 1969; Niebuhr, 2010; Qian and Stough, 2011). The more diverse a region’s population, the wider the range of available knowledge stemming from people’s various backgrounds. More available knowledge provides more ways of combining existing knowledge. Schumpeter (1934) defines innovation exactly as new combinations of existing knowledge. People with different backgrounds may also have different perspectives of thinking, which makes the carrying out of new combinations—i.e. Schumpeterian innovations—more likely to occur. In a comprehensive study of the value of diversity, Page (2007, p. 330) finds that “diverse perspectives, interpretations, heuristics, and predictive models should improve problem solving and prediction”. The nature of innovation is associated with the problem solving and prediction in the market context.
Thirdly, tolerance may facilitate knowledge spillovers (Florida et al., 2008). A knowledge spillover generally refers to the flow of knowledge between different economic agents that occurs without compensating the costs of knowledge production. It therefore reduces the social costs of innovation. Knowledge spillovers are particularly important in the context of regional development, as many studies have shown that spillover effects tend to be localised or geographically bounded (Acs et al., 2002; Almeida and Kogut, 1999; Anselin et al., 1997; Audretsch and Feldman, 1996; Jaffe, 1989; Jaffe et al., 1993; Peri, 2005). Localised knowledge spillovers depend on the extent to which face-to-face communication may occur (Storper and Venables, 2004). Tolerance or openness lowers barriers to communication among people and especially between people of different backgrounds, thus creating more opportunities for knowledge spillovers.
Florida’s empirical studies have largely supported the positive associations between tolerance and technology or innovation (Florida, 2002; Florida and Gates, 2001; Florida et al., 2012; Florida et al., 2008; Mellander and Florida, 2011).
2.2 Tolerance and Entrepreneurship
Tolerance may also contribute to entrepreneurship through three mechanisms. First, and similar to its indirect effect on innovation, tolerance (i.e. low barriers to entry) is associated with a large presence of talent in a region (Florida, 2002). Entrepreneurs themselves are a part of a region’s talent pool and may accordingly be attracted to tolerant, diverse regions. Moreover, human capital is not only one source of entrepreneurial opportunities (Acs et al., 2009; Audretsch and Lehmann, 2005), but also the major source of entrepreneurial absorptive capacity that represents the ability of the entrepreneur to understand new knowledge, recognize its value, and subsequently commercialize it by creating a firm (Qian and Acs, 2011).
While a more tolerant region ceteris paribus is associated with a higher level of human capital (Florida et al., 2008), tolerance may have a positive effect on entrepreneurship.
Secondly, diverse economic agents perceive and value potential market opportunities differently, which makes the discovery and exploitation of potential market opportunities more likely to occur (Audretsch et al., 2010; Cheng and Li, 2012; Qian et al., 2012). For instance, culture and language skills allow Chinese and Indian entrepreneurs in Silicon Valley to exploit more effectively the technology market opportunities in their home countries (Saxenian, 2002). Similarly, an entrepreneur with any unique background may attach a higher potential value to (or have advantages in discovering and exploiting) a market opportunity that pertains to her or his own culture or background. As a consequence, a region with a culturally diverse population may take advantage of the diverse types of entrepreneurial activity that can be attributable to cultural diversity.
Thirdly, population diversity may lead to the demand for diverse products and services, thus bringing diversified market opportunities that await entrepreneurs’ exploration. Porter (1995) notes that culturally or ethnically specific needs bring specific market opportunities to businesses, which reflect the inner cities’ competitive advantage. A diverse population’s demand for diverse food and restaurants is a good example of this mechanism. Porter further argues that retailing tailored for certain ethnic groups creates additional market opportunities for distributors and producers through backward linkages. In the context of regional studies, this mechanism may be more applicable for retail- or service-based start-ups that generally service a local market than high technology start-ups that are less likely to have a geographically bounded market.
On the empirical side, Audretsch et al. (2010), Cheng and Li (2012), Lee et al. (2004) and Qian et al. (2012) have found positive associations between tolerance and entrepreneurship.
3. Differentiating Tolerance and Diversity
3.1 Tolerance and Diversity: Conceptual and Empirical Differences
One of the limitations in the existing literature is that the terms tolerance, openness and diversity tend to be conflated or interchangeably used. In his ‘3Ts’ theory that addresses the importance of talent, technology and tolerance in regional economic development, Florida has not clearly distinguished these terms. He defines tolerance as “openness, inclusiveness, and diversity to all ethnicities, races, and walks of life” (2003, p. 10). He has primarily used three measures of tolerance or diversity in various studies: the Gay Index, which measures the share of the gay and lesbian population; the Bohemian Index, which measures the share of the artistic population; and the Melting Pot Index, which measures the share of the immigrant population. In The Rise of the Creative Class (Florida, 2002), the Gay Index is explicitly considered both “a good measure for diversity” (p. 255) and “a leading indicator of a place that is open and tolerant” (p. 258). The 10th anniversary edition of this book (Florida, 2012) again does not refer to the differences between tolerance and diversity, except that “openness to diversity” is considered as a “broadly speaking” replacement of “tolerance” (p. 232).
While we agree with Florida on the similarity between tolerance and openness in the context of regional development, we argue that they should be distinguished from diversity. The Merriam–Webster dictionary defines tolerance as “sympathy or indulgence for beliefs or practices differing from or conflicting with one’s own” or “the allowable deviation from a standard”; in contrast, diversity is understood as “the condition of having or being composed of differing elements”, especially “the inclusion of different types of people (as people of different races or cultures) in a group or organization”. Based on these definitions, tolerance and social or cultural diversity may differ at least in two ways. First, tolerance presumes an individual standard; diversity does not. The extent to which deviation from the standard is accepted reflects how tolerant an individual is. Florida et al. (2008) cite the work of Inglehart (Inglehart and Norris, 2003; Inglehart and Welzel, 2005) who states that openness to gays and lesbians is the best indicator of tolerance. Florida et al. further construct a Tolerance Index, which is calculated by combining the Gay Index and the Bohemian Index. Diversity, on the other hand, concerns the inclusiveness of different groups and does not necessarily involve a standard. In other words, all groups can be treated equally in measuring diversity. Secondly, a higher level of tolerance does not necessarily lead to a higher level of diversity. As Ottaviano and Peri (2006) state, a high level of social or cultural diversity requires both a large number of social or cultural groups and an even distribution of individuals across these groups. Tolerance, in contrast, may be maximised simply with the co-existence and integration of the two most different groups.
The Gay Index and the Bohemian Index might be good indicators for tolerance, as argued by Florida and Inglehart, but they are not good indicators for social or cultural diversity since they cannot capture the distribution of individuals across cultural groups. Indeed, several studies have adopted straightforward measures for diversity (Audretsch et al., 2010; Cheng and Li, 2012; Ottaviano and Peri, 2006; Qian and Stough, 2011). One of the most commonly used measures of diversity is based on the Herfindahl–Hirschman index (HHI). Mathematically, it can be written as
where si represents the share of population in cultural group i; and N is the total number of groups under consideration.
This index has the largest value of 1 when the overall population belongs to one group and accordingly diversity is minimised. It has the smallest value of 1/N when each cultural group carries the same share of population and thus diversity is maximised. A Diversity Index is generally the inverse Herfindahl–Hirschman index, allowing a higher value to represent a higher level of diversity. Mathematically, the Diversity Index can be written as
which has a value of 0 with minimum diversity and a value of (N−1)/N with maximum diversity. This index, as Ottaviano and Peri (2006) note, can capture both the number of groups and the population distribution of different groups. Both of these two elements contribute to diversity.
To provide some empirical evidence on the relationship between diversity and tolerance, Figure 1 plots the Tolerance Index developed by Florida et al. (2008)—the average of the Gay Index and the Bohemian Index—against a Diversity Index that is similar to Ottaviano and Peri’s (2006) for 276 US metropolitan statistical areas (MSAs), using the data from 2000. 1 Ottaviano and Peri’s Diversity Index is calculated using equation (2) and cultural groups are defined based on countries of birth. Countries of birth may be a better reflection of cultural differences than races or ethnicities. In this study, a country of birth is used to identify one cultural group only if there are at least 100 000 people in the US that were born in the country; the populations born in all countries that do not meet this criterion are aggregated and considered as one cultural group (called “Other countries total”). In all, 51 cultural groups are eventually identified, with those born in the US making up the largest group and those born in the Mexico making up the second-largest. 2 For our sample, the Tolerance Index and the Diversity Index exhibit a positive association, with a correlation coefficient of 0.52. This coefficient suggests that, while tolerance and diversity do share similarities, each also has a large share (almost one half) of their variance that does not overlap with the other. For individual cities, San Francisco ranks the highest in tolerance, significantly outpacing second-ranked Santa Fe. However, San Francisco is not among the top five most diverse cities and Santa Fe only shows a moderate level of diversity. Meanwhile, the cities that present the highest levels of diversity, such as Miami and Jersey City, do not rank high for tolerance. The upper-left box area in Figure 1 shows that quite a few barely or moderately tolerant cities demonstrate high levels of diversity.

Scatter plot: diversity versus tolerance. Data sources: tolerance data are provided by Kevin Stolarick from University of Toronto’s Martin Prosperity Institute; diversity data are calculated based on countries of birth data from the 5 per cent American Community Survey (2000).
3.2 Innovation and Entrepreneurship: The Roles of Tolerance and Diversity Revisited
Having established the differences between tolerance and diversity, it is important to have a closer look at their distinct impacts on innovation and entrepreneurship. Section 2 has identified three mechanisms by which tolerance or diversity may affect innovation: talent attraction; diverse knowledge and diverse perspectives of thinking; and increased communication and knowledge spillovers. It also discusses three mechanisms by which tolerance or diversity may affect entrepreneurship: talent attraction; diversity in perceiving potential market opportunities; and diversity in consumer demand. Since we have conceptually separated diversity and tolerance, it is important to re-examine whether these mechanisms are primarily associated with tolerance or diversity. This effort allows for a better understanding of the social drivers of innovative and entrepreneurial activities. The differentiated roles of tolerance and diversity in innovation and entrepreneurship are summarised in Table 1.
Social drivers of innovation and entrepreneurship: tolerance and diversity differentiated
In the case of innovation, the first mechanism is primarily about talent attraction. The literature reviewed earlier suggests that both tolerance and diversity matter in attracting and retaining human capital. Tolerance is relevant because it represents low barriers to entry; diversity is relevant because it indicates diverse cultures, foods and entertainment that make a place attractive to talent. The second mechanism revolves primarily around diversity that allows for great variations in knowledge and perspectives of thinking, and has little to do with tolerance. In contrast, the third mechanism represents the role of tolerance (i.e. openness to communication with others) in facilitating knowledge spillovers and has little to do with diversity.
In the case of entrepreneurship, the first mechanism is the same as the case of innovation—i.e. via talent attraction—and again involves both tolerance and diversity. The second and third mechanisms, which address diverse ways of perceiving market opportunities and diversified consumer demand respectively, are primarily associated with diversity and have little to do with tolerance.
4. The Social Driver of Innovation and Entrepreneurship: Diversity versus Tolerance
This section compares the effects of tolerance and diversity on innovation and entrepreneurship. In the previous section, the mechanisms by which differentiated tolerance and diversity may contribute to innovative and entrepreneurial activities have been discussed. This section provides empirical evidence of these mechanisms. Following many regional studies of innovation and entrepreneurship (Anselin et al., 1997; Florida et al., 2008; Glaeser et al., 2010; Lee et al., 2004; Qian et al., 2012), this study uses US MSAs as the geographical unit for analysis. The effects of tolerance and diversity on innovation and entrepreneurship are examined in multivariate analysis.
4.1 Variables, Model and Data
Three dependent variables are separately used: innovation, entrepreneurship and high technology entrepreneurship. High technology entrepreneurship is separated from entrepreneurship and highlighted because the regional development effects of tolerance and diversity are mostly discussed in the context of knowledge economies. The number of patents per 1000 inhabitants is employed to measure innovation. Despite not being an ideal proxy for innovation (Griliches, 1990), patent information is largely used in regional studies (Acs et al., 2002; Jaffe, 1989; Peri, 2005). Following Qian et al. (2012), entrepreneurship is measured by the new firm formation rate, which is the number of new single-unit establishments divided by the labour force (in 1000s). High technology entrepreneurship is measured by dividing the number of new single-unit establishments in high technology industries 3 by the labour force (in 1000s). New single-unit establishment data by county and by industry are available from the US Census Bureau’s Business Information Tracking Series (BITS). The county-level data are aggregated into MSA-level data.
The primary explanatory variables are tolerance and diversity. As discussed in the previous section, tolerance is measured by the average of the Gay Index and the Bohemian Index, and diversity is the inverse Herfindahl–Hirschman index based on countries of birth data.
According to the literature, some variables may be associated with both innovation and entrepreneurship. First, in the context of new growth theory, Romer (1990) considers human capital as an input of knowledge or innovative output. A large body of empirical research at the regional level—for example, Faggian and McCann (2009)—supports the contribution of human capital to innovation. Human capital is also associated with entrepreneurship. It is an indicator of new knowledge that represents entrepreneurial opportunities (Acs et al., 2009; Audretsch and Lehmann, 2005). In addition, human capital is the major source of the entrepreneurial absorptive capacity that allows entrepreneurs to exploit successfully new knowledge via creating new firms (Qian and Acs, 2011). Empirical studies by Acs and Armington (2006) and Qian et al. (2012) support the positive effect of human capital on entrepreneurship.
Secondly, the role of universities in innovation and entrepreneurship has been widely recognised. University R&D represents one of the major sources of innovation and is one of the major inputs of the knowledge production function (Acs et al., 2002; Anselin et al., 1997; Jaffe, 1989). Universities may contribute to local entrepreneurial activity via knowledge spillovers and technology transfers (Bercovitz and Feldman, 2006), business training (Katz, 2003), university-affiliated incubators (Qian et al., 2011) and programmes that provide incentives for faculty/student start-ups (Smilor et al., 2007).
Thirdly, the innovative, entrepreneurial economy in some high technology regions (such as Silicon Valley) suggests the impact of technology. Feldman and Florida (1994) address the importance of technological infrastructure to innovation. Florida (2012) shows substantial increases of innovative activity in high technology regions from 1976 to 2009. Some work on Silicon Valley has highlighted high technology entrepreneurship (Harrison et al., 2004; Kenney and von Burg, 1999).
Lastly, agglomeration has gained much attention as a factor in the knowledge-based regional economy. According to Marshall (1920) and Porter (1998), agglomeration or geographical proximity facilitates knowledge spillovers that are important to both innovation (Jaffe, 1989) and entrepreneurship (Acs et al., 2009; Audretsch and Lehmann, 2005). On the empirical side, Carlino et al. (2007) find that urban employment density is positively associated with patent activity. Delgado et al. (2010) find a positive effect of clusters on entrepreneurship.
Based on the literature, this analysis controls for the effects of human capital, universities, technology and agglomeration. The regression model can be written as
where, the dependent variable innovation can be replaced by entrepreneurship or high technology entrepreneurship. The share of adult population (i.e. age 25+) with a bachelor’s degree is used here to measure human capital. The presence of universities is measured by the number of university faculty per 1000 inhabitants. The Milken Tech Pole Index (DeVol, 1999), an index measuring the concentration of high technology industries, is employed as a proxy for technology. Finally, the natural log of population per square mile is used to measure agglomeration. By using population density, the agglomeration variable captures urban agglomeration but not industrial agglomeration.
The model in equation (3), however, does not allow for testing the first mechanism (i.e. talent attraction and retention) by which tolerance and diversity contribute to innovation and entrepreneurship. This is because the mediating factor—human capital (i.e. talent)—is also included among explanatory variables, which cancels out the talent attraction effects of tolerance and diversity. To remedy this, we run additional regressions using tolerance and diversity as explanatory variables of human capital. If the empirical results support a positive effect of human capital on innovation or entrepreneurship and meanwhile support a positive effect of diversity or tolerance on human capital, the indirect role of diversity or tolerance on innovation or entrepreneurship via talent attraction will be supported. In addition to tolerance and diversity, the literature identifies the presence of universities as a predictor of the region’s human capital (Abel and Deitz, 2012; Berry and Glaeser, 2005; Florida et al., 2008). Moreover, amenities or the quality of life has also been found as an attractor of talent (Florida et al., 2008; Glaeser et al., 2001; Wenting et al., 2010). Accordingly, the human capital model being tested is
Equation (4) is similar to the talent model in Florida et al. (2008). The difference here is that diversity and tolerance are separated. The measures for tolerance, diversity and the university remain the same as previously discussed. Following Florida et al. (2008), amenities or the quality of life is measured by the number of five-digit NAICS service industries that were present in the region in 2000. 4
As for the time-period of data, the 2000 data are used for all independent variables in this cross-sectional analysis. The 2001 patent data are used for the innovation variable and the 2002/03 new firm formation data are used for the entrepreneurship variable. Time lags are used to address partially the possible effect of endogeniety. Out of all 331 US MSAs, 276 MSAs are included in the final sample. 5 Table 2 presents the descriptive statistics of all variables.
Descriptive statistics
Kevin Stolarick provides cleaned data for this variable.
Table 3 is the correlation matrix. In a bivariate context, both tolerance and diversity are positively, significantly associated with entrepreneurship, high technology entrepreneurship and innovation. Among all explanatory variables, besides the coefficient between diversity and tolerance (with a value of 0.52, as previously discussed), human capital and tolerance exhibit a strong correlation with a coefficient value of 0.71. This echoes Florida’s argument (2002) on the strong relationship between tolerance and talent attraction.
Correlation matrix
Note: ** significant at 0.05.
4.2 Findings
Tables 4, 5 and 6 show OLS regression results with innovation, entrepreneurship and high technology entrepreneurship as separate dependent variables. In each case, tolerance and diversity are included in the regressions first individually and then jointly. Therefore, three models are tested for each dependent variable. 6
Regression results: innovation as dependent variable
Notes: *** significant at 0.01; ** significant at 0.05.
Regression results: entrepreneurship as dependent variable
Notes: *** significant at 0.01; ** significant at 0.05.
Regression results: high technology entrepreneurship as dependent variable
Notes: *** significant at 0.01; ** significant at 0.05.
For our primary variables of interest, Table 4 reveals that neither tolerance nor diversity has a significant effect on innovation. This is inconsistent with the findings of Florida et al. (2008) and Qian and Stough (2011). These results do not support the second mechanism (i.e. diverse knowledge backgrounds and diverse perspectives of thinking) by which diversity contributes to innovation and the third mechanism (i.e. increased communication and knowledge spillovers) by which tolerance contributes to innovation. This may be explained by the existence of high communication costs between culturally different groups (Lazear, 1999) and a lack of social cohesion in diverse communities (Putnam, 2007). Table 5 shows that both tolerance and diversity present a positive, significant association with entrepreneurship when each appears in the regression alone. However, when both are included, it is diversity that remains positive and significant. The effect of tolerance, despite remaining positive, becomes insignificant. The same pattern holds for high technology entrepreneurship models, as suggested in Table 6. These results support the second mechanism (i.e. diverse ways of perceiving market opportunities) and/or the third mechanism (i.e. diversified market demand) by which diversity contributes to entrepreneurship.
For the control variables, Table 4 shows that human capital is positively, significantly associated with innovation, echoing Romer’s knowledge production function (1990) that uses human capital as an input for innovative output. In Tables 5 and 6, the coefficient of human capital is consistently positive and significant in all entrepreneurship and high technology entrepreneurship models. This suggests the important role of human capital in entrepreneurial economies, consistent with the empirical findings of Acs and Armington (2006), Audretsch and Lehmann (2005), Lee et al. (2004) and Qian and Acs (2011).
Not surprisingly, technology is positively, significantly associated with innovation, as demonstrated in Table 4. Despite the popularity of high technology entrepreneurship among scholars (for example, Harrison et al., 2004) and economic development practitioners, technology presents a negative association with entrepreneurship. This negative effect is significant in two of the three entrepreneurship models in Table 5. A positive, significant effect of technology is even barely seen in high technology entrepreneurship models in Table 6. Although Silicon Valley as the US high technology centre is highly entrepreneurial, most start-ups in the US have little to do with high technology industries: Table 2 shows a much higher mean of entrepreneurship (4.32) than that of high technology entrepreneurship (0.30).
Surprisingly, ceteris paribus, the university variable is negatively associated with innovation, entrepreneurship and high technology entrepreneurship. Agglomeration is only positively associated with high technology entrepreneurship and this relationship is not significant. The relationships between university/agglomeration and innovation/entrepreneurship, although not the focus of this research, deserve further exploration in future research.
The regression results in Table 7 show that both tolerance and diversity exhibit a positive, significant association with human capital when only one of them is included in the model. The adjusted R 2 of model 1 (in which tolerance is an explanatory variable) is much larger than that of model 2 (in which diversity is an explanatory variable), suggesting tremendous explanatory power of tolerance in the geographical concentration of human capital. When both are present as explanatory variables, the coefficient of tolerance remains positive and significant, and the coefficient of diversity becomes negative and significant at the 0.1 level. The results show that it is tolerance (but not necessarily diversity) that plays an important role in talent attraction and/or retention. And talent or human capital, according to the results in Table 4, Table 5 and Table 6, further contributes to both innovation and entrepreneurship. Therefore, the first mechanism listed in Table 1 is partially supported. Tolerance has an indirect effect on innovation and entrepreneurship via talent attraction, consistent with Florida (2002).
Regression results: human capital as dependent variable
Notes: *** significant at 0.01; ** significant at 0.05.
Both the quality of life and the university are positively associated with human capital, which is consistent with the literature. These two variables therefore may also have indirect effects on innovation and entrepreneurship via human capital.
5. Summary and Discussion
The work of Richard Florida (2002, 2012) highlights the role of tolerance and diversity in regional and urban economic development, which has had a tremendous influence on urban planning research and practice. Surprisingly, diversity and tolerance have not been very clearly defined and in most research they are conflated or interchangeably used. This article argues that diversity and tolerance are conceptually different and that it is important to distinguish one from the other. This distinction allows us to reach a better understanding of the separate roles of tolerance and diversity in regional and urban development. Tolerance involves individual standards; diversity does not. Openness to greater deviation from one’s own standard represents a greater level of tolerance; a shift towards the even distribution of population across different social or cultural groups increases diversity. A correlation coefficient of 0.52 between tolerance and diversity derived from our sample, measures and data suggests that they are also empirically different.
Our efforts to differentiate diversity and tolerance may lead us to rethink the mechanisms by which these two factors contribute to innovation and entrepreneurship in the urban context. To begin with, we have found a positive association between tolerance (but not necessarily diversity) and human capital in a multivariate context, and human capital further demonstrates a positive effect on both innovation and entrepreneurship. Therefore, tolerance matters for innovation and entrepreneurship in an indirect way: it signals low barriers to entry and accordingly helps to attract and retain innovative or entrepreneurial talent, a major argument made by Florida (2002).
In terms of direct effects, diversity has a positive, significant association with entrepreneurship, both in the high technology context and for the overall regional economy. These results suggest that diversity indeed provides diversity in perceiving entrepreneurial opportunities (i.e. the second mechanism in the case of entrepreneurship) and/or diversity in consumer demand or entrepreneurial opportunities (i.e. the third mechanism in the case of entrepreneurship).
However, diversity has no significant association with innovation, meaning that the role of diversity in innovation via diversifying knowledge and perspectives of thinking (i.e. the second mechanism in the case of innovation) is not supported in our empirical analysis. This may be explained by high communication costs between different social or cultural groups, which could create barriers to knowledge spillovers and the carrying out of new combinations. As Lazear (1999) has noted, homogeneous culture associated with shared customs and mutual trust may facilitate trades and communication among individuals. Diversity in language and culture, in contrast, may negatively impact the willingness to communicate. Putman (2007) notes that “inhabitants of diverse communities tend to withdraw from collective life, to distrust their neighbours” (p. 150) and “ regardless of the colour of their skin, to withdraw even from close friends” (pp. 150–151). Lazear suggests that the overall effect of diversity depends on the trade-off between diversity’s productivity benefits and the communication costs it may create.
After isolating its indirect effect via talent attraction, tolerance exerts no significant impact on both innovation and entrepreneurship. Compared with the theoretical framework in Table 1, the result does not support the third mechanism in the case of innovation—i.e. by facilitating communication and knowledge spillovers. One social group’s tolerance for the presence of another social group with a different value standard, sexual orientation or lifestyle in the same region does not necessarily mean they are socially cohesive and willing to communicate with each other. Indeed, even within the same metropolitan area, inhabitants with the same identity or culture tend to agglomerate in certain communities, leading to segregation and reduced interaction among different groups (Blair and Carroll, 2009). For instance, in San Francisco, the most tolerant city in our sample, to build spatially concentrated gay communities was “a fundamental characteristic of the gay liberation movement” (Castells, 1984, p. 138).
Our empirical analysis calls for urban policy to remove the barriers to communication and facilitate social cohesion among culturally diverse groups, so that diversity and tolerance can play their expected role in innovative activity. A large body of empirical research shows that diversity is associated with eroded trust (Rothwell, 2012) and may thus discourage communication and collaboration. As revealed in recent literature, however, it is segregation or the lack of interaction among spatially bounded social groups, but not diversity itself, that accounts for the decline of trust or social capital (Rothwell, 2012; Stolle et al., 2008; Uslaner, 2011). Public policy to address the segregation problem will have positive spillover effects on innovation in cities.
6. Future Research
This research may spur more efforts to reach a better understanding of the different roles of tolerance and diversity in regional and urban development. According to Fainstein (2005, p. 3), diversity has become “the new orthodoxy for city planning”. However, this principle may shed little light on urban public policy without clear definitions of diversity and tolerance and without understanding their unique impacts on regional and urban development. Research in the immediate future may include the use of additional measures of diversity and tolerance, the improvement of model specifications and the exploration of the impacts of diversity and tolerance on other aspects of regional development. The measures of tolerance and diversity in this research are primarily built upon Florida’s focus on gays, bohemians and immigrants. When one refers to social diversity, however, the racial perspective is typically very important but is not considered in this study. In contrast to the positive effects of cultural diversity measured through immigrants on regional economic development (Autretsch et al., 2010; Niebuhr, 2010; Ottaviano and Peri, 2006; Qian and Stough, 2011), racial diversity is generally found to be not associated with innovation, entrepreneurship or economic development in general (Cheng and Li, 2012; Florida, 2012; Page, 2007). In addition to measurement issues, model specifications may need additional work, since current models’ explanatory power, as suggested by the adjusted R 2s in Tables 3 and 4, is relatively weak in some cases. Lastly, it is also interesting to examine the unique contributions of differentiated diversity and tolerance on other aspects of urban development, such as productivity, income and housing. In all, comprehensive, systematic investigations of tolerance and diversity in the context of urban development are still needed. These efforts would allow urban planners to reach a better understanding of the social underpinnings of urban economic development.
Footnotes
Acknowledgements
The author would like to thank Kevin Stolarick for his data support and Minkyu Yeom for his research assistance. Access to the BITS data was made possible by the School of Public Policy, George Mason University. The final version of this paper benefits greatly from the comments made by Editor Andrejs Skaburskis and three anonymous referees of Urban Studies, as well as the participants in the 2012 “Experience the Creative Economy” Conference held at the University of Toronto. All errors are the author’s own.
Funding
This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
