Location patterns and determinants of knowledge‐intensive industries in the Tokyo Metropolitan Area

Abstract

The knowledge industry has an important function for regional economic development in cities and countries with strong knowledge economies. This study examined the characteristics of clustered Knowledge Intensive Business Services (KIBS) in the Tokyo Metropolitan Area and investigated the determinants of these enterprises’ locations. Using municipal‐level data, the significance of the determinants of KIBS’ locations was tested in a path analysis, followed by analysis of questionnaire survey data on KIBS enterprises. The results were as follows. First, several dense KIBS agglomerated clusters were found in central Tokyo and some inner‐suburban business districts. Ellison‐Glaeser index statistics confirmed that KIBS were more likely than non‐KIBS to geographically concentrate. Second, the most influential determinants of KIBS’ location were urban economies, food amenities, and nightlife amenities. The result of the questionnaire survey also implies that, in highly urbanized areas, a dense distribution of business enterprises, food and nightlife amenities, and sophisticated urban regional image creates a positive urban environment evaluation by KIBS enterprises and attracts them through direct and indirect causal relationships.

Introduction and background

In many highly developed countries, the manufacturing sectors have become less prominent, and the importance of the knowledge economy is increasing . The Establishment and Enterprise Census of Japan (EEC) reported that the Tokyo Metropolitan Area (TMA), which is Japan's capital region, has experienced a continuous increase in the knowledge industry's workforce since the decline of manufacturing workers that began after it peaked in 1991. In the TMA and similar cities, the knowledge industry is a sector of the economy primarily involved in creating knowledge, and it plays an important part in regional economic development.

The knowledge industry has unique characteristics in terms of its location because, for example, it features the easy and low-cost transmission of intermediate and final products, little need for logistical functions, a large need for face-to-face communications, and so on. Therefore, it is assumed that, along with the shift to a post-industrial society, the locations of business enterprises in metropolitan areas will significantly change. Consequently, it is important to study and understand how enterprises’ locations in metropolitan areas are changing in response to the dramatic changes occurring in the industrial structure.

Debate continues about the way to quantitatively define and understand the knowledge industry. This study focused on Knowledge Intensive Business Services (KIBS), which create and provide advanced knowledge to businesses. KIBS is one of the fastest growing segments of the knowledge industry, and it is recognized as an important industrial segment driving overall regional economic development and as having a significant role in regional technological innovation . Therefore, this study identified KIBS as a new driving force of urban spatial development and structural transformation in the post-industrial era. This study found changes in the reasons for enterprise locations by comparing KIBS’ and non-KIBS’ locations (general service industries). This study aimed to achieve the following 2 objectives:

1. Identifying the locations of KIBS in the TMA and their agglomeration density compared to non-KIBS.
2. Identifying the intra-metropolitan determinants of KIBS’ locations in the TMA.

This study defined “KIBS” using the Japan Standard Industry Classifications (JSIC), which is explained in Section . Sections  to present the study's analysis of the KIBS spatial distribution and agglomeration density in the TMA compared to that of non-KIBS. Section  discusses previous studies on the locations of knowledge industry enterprises and extracts from them the possible determinants of location examined by this study. Then, indicators were developed for each determinant for quantitative analysis, and a hypothetical location model was created based on these indicators. The dataset based on the location model is described in Section , and the results of a path analysis are explained in Section . This step validated the hypothetical location model and tested the significance of each indicator as a KIBS’ location determinant. This step was repeated on the non-KIBS data, whose results were then compared to those of the KIBS, and the relative location characteristics were analyzed. Based on these results, Sections  to explain several significant factors determining KIBS’ location and report on a questionnaire survey conducted on KIBS enterprises. Through a questionnaire survey of KIBS executives, understanding was deepened on the actual location preferences of KIBS enterprises. Chapter 7 summarizes the results and offers conclusions.

Literature review

Previous empirical studies on the knowledge industry's reasons for certain locations have mostly been set in Western cities, and studies on Japanese cities are rare. The major findings of the previous studies are summarized as follows.

Many previous studies have noted that the knowledge industry tends to be attracted to large cities, suggesting that a location's economic scale is one of the most important determinants of location choice. Shearmur et al. analyzed the spatial distribution of Canada's knowledge industry and found that it was primarily located in large cities. Stam conducted an empirical study on the locations of small and mid-sized high-growth knowledge enterprises in the Netherlands, also finding them to be concentrated in city centers. For Norway's software industrial agglomeration in Oslo, Isaksen analyzed the location determinants using survey data and clarified the importance of proximity to customers and peer enterprises.

Saxenian and Florida focused on the relationship between the locations of high-tech industries and their local social environments . Florida developed sets of indicators on the social environment as possible location determinants of high-tech industries and quantitatively tested causal relationships. The study found that tolerance and geographical distribution of talent significantly influenced areas’ abilities to attract high-tech enterprises. Florida also analyzed the effects of various regional amenities, such as cultural amenities and recreational amenities, on high-tech industries’ locations, but he found no conclusive causal relationships.

Only a few studies on knowledge industry locations have been conducted in Japan. Kwon et al. clarified the location determinants of the knowledge service industry at the prefectural scale, but the current study is the first empirical analysis of KIBS’ location mechanisms at the intra-metropolitan scale.

This study is the first detailed empirical analysis of the KIBS’ location mechanisms in Japan's capital region at the metropolitan level. The association between KIBS’ locations and their socioeconomic and physical contexts is examined to improve understanding of the reasons that these enterprises form clusters in the TMA. Municipal-level data were obtained, prepared, and analyzed to describe the socioeconomic and physical contexts of the KIBS’ locations. The study is characterized as a geographically detailed, intra-metropolitan study using municipal-level data that explains the relative characteristics of KIBS’ locations through comparisons to non-KIBS enterprises.

Methods Study site and sample

This study's units of analysis were KIBS business enterprises in the TMA (Figure ) . Some areas of the TMA had small absolute numbers of enterprises, particularly at the TMA's periphery. Some of these areas contained small numbers of large-scale enterprises, such as a huge suburban workplace of a financial service company with many clerical workers, which created a risk that the study's measure of KIBS’ location tendency (the Location Quotient) would be erroneously high. Consequently, wide variation in the enterprises’ sizes risked instability of the analytical results when location was measured at the municipal level. Therefore, to secure analytical stability, only business enterprises with less than 100 workers were included .

There is no scholarly consensus on the definition of KIBS, although a concrete definition is crucial for empirical analyses of KIBS’ locations. Miles defined KIBS as “Services that provide knowledge-intensive inputs to the business processes of other organizations”. The current study followed Miles and identified an enterprise as a KIBS when it satisfied the following two criteria: (i) its main goal was to provide services to other business enterprises and (ii) its main function was to produce and introduce knowledge. Using these two criteria, KIBS were identified using the JSIC .

First, “business services” were identified using the JSIC 2-digit industrial classifications. Twelve classes were extracted from them of enterprises that mainly performed support activities for other enterprises . Then, detailed features of these 12 classes were investigated using a 3-digit classification. Based on the detailed findings, 5-three-digit classes were dropped because they were expected to have a relatively large personal service component .

From among the 3-digit JSIC “business services” classes, enterprises whose “main task was production and introduction of knowledge” were extracted after first excluding the 3-digit classes with ambiguous definitions that made it difficult to determine whether they satisfied the criteria of knowledge intensiveness. The excluded classes were “other professional service industries,” “other technical service industries,” and “administrative or ancillary economic activities.” Finally, 20-three-digit JSIC classes were extracted as KIBS, comprising the study sample (Table ). Therefore, this study defined KIBS as these 20 JSIC classes.

JSIC classes extracted as KIBS

391	Computer Programming & Other Software Services
392	Data Processing & Information Services
401	Services Incidental to Internet
742	Engineering & Architectural Services
743	Mechanical Design Services
411	Video Picture Information Production & Distribution
412	Sound Information Production
413	Newspaper Publishers
414	Publishers, Except Newspapers
415	Commercial Art & Graphic Design
416	Services Incidental to Video Picture, Sound, Character Information Production and Distribution
721	Lawyers’ & Patent Attorney's Offices
722	Notaries Public's, Judicial Scriveners’& House Surveyors’ Offices
723	Administrative Scriveners’ Offices
724	Certified Public Accountants’ & Certified Tax Accountants’ Offices
725	Certified Social Insurance & Labor Consultants’ Offices
726	Design Services
727	Authors & Artists
728	Business Consultants & Pure Holding Companies
731	Advertising

The Location Quotients (LQ) measure of spatial distribution

To identify the characteristics of KIBS spatial distributions, the LQ values of each municipality were computed according to the number of KIBS and non-KIBS enterprises’ employees , . Municipal scale data of ECC (Number of employees in every 2-digit industrial classification and nighttime population) were obtained from the website of the National Bureau of Statistics. Figures  and illustrate the relationships of each municipality's LQ to its distance from central Tokyo. The “distance from central Tokyo” was determined using a straight line between the Tokyo station and the geographical center of each municipality.

Analytical methods

To assess the distributions of KIBS and non-KIBS in the TMA, the LQ values were assessed and compared. To quantitatively determine the agglomeration densities of KIBS and non-KIBS, the Ellison-Glaeser (EG) index was employed. The EG is widely used to measure and compare agglomeration across industries. Its basic purpose is to (i) quantify the extent to which the location pattern of a particular industry deviates from the average location pattern of all industries based on employee spatial distribution data and (ii) deduct from it the effect of the disparity in enterprise size based on individual-level enterprise data on their number of employees.

Spatial distribution of KIBS in the TMA KIBS and non-KIBS distributions in the TMA

The LQ values of the non-KIBS were mainly distributed between 0.5 and 1.5, and no clear relationships were observed between non-KIBS’ locations and their distances from central Tokyo. However, most of the KIBS LQs in many locations were between zero and 0.5, while there are several places where KIBS is strongly agglomerated with the LQ exceeding 2.0. KIBS were particularly concentrated near central Tokyo, suggesting that KIBS’ locations tended toward dense agglomeration near city centers, which was a clear contrast to the dispersed locations of the non-KIBS.

Next, the LQs were plotted on the map to geographically visualize the KIBS’ and non-KIBS’ locations (Figure ). The results visually confirm that KIBS were more densely agglomerated than non-KIBS, and several distinct KIBS clusters were observed in central Tokyo and some nearby suburbs. Focusing on the municipalities with KIBS at LQ of 1.0 or higher, the map shows that four areas could be identified as major KIBS agglomeration areas in the TMA: (i) central Tokyo (Shibuya Ward, Chiyoda Ward, Shinjuku Ward, Minato Ward, Chuo Ward, Bunkyo Ward, Toshima Ward, Nakano Ward, Taito Ward, Meguro Ward, Shinagawa Ward, Suginami Ward) and Musashino City; (ii) central Yokohama (Naka Ward, Kanagawa Ward, Nishi Ward, and Kohoku Ward); (iii) along the Chuo commuter train route (Kokubunji City and Tachikawa City); and (iv) Urawa Ward of Saitama City.

Analytical equations on agglomeration density

As the first step to compute the Ellison-Glaeser (EG) index, through Equation 1, the simple deviation of the “location pattern of a particular industry from the average location pattern” was expressed by the sum of squares of the “regional share gap” between the KIBS and all industries of the TMA.[Image Omitted. See PDF]where, Xai: Share of region-A's employment in industry-I; Xa: Share of region-A's employment in all industries.

When the Gi of an industry is high, the spatial distribution of that industry strongly deviates from the average location pattern, which implies dense agglomeration of that industry in that location. However, measuring industrial agglomeration only with the Gi should be done with caution because the “dispersion of enterprise scales,” which influences the Gi, varies among industries. If, in a given area, employees of a particular industry were employed by one or a few particularly large business enterprises, the Gi value easily might be high. However, the actual agglomeration density would not align with the meaning of “dense” agglomeration, defined as “many business enterprises in close proximity to each other.” To avoid the problem, the EG excludes the influence of a few particularly large enterprises by expressing the dispersion of the enterprises’ sizes and incorporating that information into the model. This procedure was as follows.

“Agglomeration of employment in all industries” (S) and “agglomeration of business establishments in industry-I” (Hi) were defined by Equations 2 and 3, respectively.[Image Omitted. See PDF][Image Omitted. See PDF]where, Zki: Share of establishment-K's employment in industry-I;

The Gi was divided by the “agglomeration of employment in all industries” (S) variable to obtain Gi′ as shown in Equation 4.[Image Omitted. See PDF]

The expected value of Gi′ was expressed through Equation 5 as a dichotomy comprising “agglomeration of enterprises” and “other.” In Equation 5, γi indicates the extent of agglomeration density remaining after excluding the influence of “agglomeration of enterprises.”[Image Omitted. See PDF]

When Equation 5 is solved for γi, the EG is obtained. [Image Omitted. See PDF]

To calculate the EG, it is necessary to obtain Hi. Additionally, to obtain Hi, individual-level enterprise data are needed on their sizes within their industries. However, the EEC does not provide these data on enterprises. Instead, it provides data on the “number of enterprises and the average number of employees per establishment” on individual municipalities aggregated into eight establishment size categories. Hi was approximated by Equation 7 using these data.[Image Omitted. See PDF]where, r: 8 establishment size categories in ECC; Wri: Average number of employees per establishment in industry-I within ECC establishment size category-R; Wi: Number of employment in industry-I.

Agglomeration density results

The 2009 EEC dataset was employed to compute Equation 7 to estimate values of Hi. Then, by substituting the value in Equation 6, γi for each of the KIBS and non-KIBS was calculated, which derived the EG on the KIBS (0.023562) and the EG on the non-KIBS (0.000148). A large EG implies that the agglomeration is relatively dense. The results of the EG comparison indicate that the KIBS EG is larger than that of the non-KIBS, confirming that KIBS were more densely agglomerated than non-KIBS, which are the less knowledge-oriented service industries.

Determinants of knowledge industries’ locations Possible location determinants

To quantitatively investigate the location determinants of the KIBS enterprises, it was necessary to include as many measures as possible of all the predictive factors. However, not all of the reasons for location choices made by KIBS are known, and there is a lack of consensus about the known determinants. Therefore, known determinants are in dispute, and unknown determinants likely exist. The location determinants used in this study's analysis were found through a literature review and assessment and by listing location selection attributes (of all business enterprises including KIBS) identified by the survey results of the Real Estate White Paper. Using these data, location determinants relevant to this study were chosen to represent both data sources, select measures for each factor, and develop a hypothetical location model for statistical testing.

The regional economy

Shearmur et al., Stam, and Isaksen each found that the economic scale of a region is an important determinant of knowledge industries’ locations. These findings imply that knowledge industries are strongly attracted to urban economies for the benefits of geographical concentration. The previous studies stated above suggested that knowledge industries might be more focused than other types of industries on urban economies, which is important to address and which was included in the current study's analysis (Figure ).

The social environment

Florida and Saxenian investigated the causal relationships between environmental factors and high-tech industrial locations. Their results suggest that the quality of the social environment, such as the extent of “tolerance” and the “geographical distribution of talent,” might directly and indirectly influence where knowledge industries locate. Therefore, the social environment was measured and tested in this study (Figure ).

Urban amenities

Florida also examined the relationship between high-tech industrial locations and urban amenities, hypothesizing that regional urban amenities positively influenced the geographical distribution of talented people and attracted more high-tech industries, but he found no statistically significant relationships from the path analysis (p. 118). However, he noted that the lack of statistically significant results likely was because his measures of cultural and recreational amenities did not function well (p. 114). However, the qualitative analysis of Florida's survey interview data found that talented employees tended to be attracted to certain areas for their rich urban amenities (p. 95). Therefore, the current study tested the effects of regional amenities on the KIBS’ locations and the effects of regional amenities on the geographical distribution of talented people, which, in turn, might relate to the KIBS’ locations (Figure ).

Location selection based on Real Estate White Paper data

The Real Estate White Paper is an annual publication of Ikoma Data Service System, Ltd., which is a private real estate firm that is one of the most comprehensive sources of workplace market data in Japan. In 2006, it conducted a questionnaire survey on workplace selection among private enterprises. It found 11 categories of workplace selection criteria. Nine of these criteria were used in the current study because they are supposed to influence enterprise location at the municipal level, which is the level of analysis of this study . Figure  shows the proportional distribution of the variables that private enterprises in the TMA regard as “important to workplace selection,” from the survey results of the Real Estate White Paper.

The 9 “important factors” for workplace selection mentioned in Figure  are classified into 5 categories. The following sentences provide the details of these factors:

1. Regional image concerns the perception of the influence of a region's reputation on an enterprise's and its employees’ reputations.
2. Access to other enterprises, such as business partners, parent and affiliated enterprises, and enterprise-related facilities, influences the communication costs (time and money for telephonic and postal communications and travel for face-to-face interaction) and the density of mutual learning.
3. Employee transportation influences enterprises because it determines the employees’ costs of travel to the workplace (in many Japanese enterprises, employees’ commuting expenses are paid by the employers). Employees are also directly influenced because it determines their costs in time spent commuting.
4. Access to everyday resources concerns access to goods and services necessary for an enterprise to operate and that are used on a daily basis, such as office supplies, banks, and post offices.
5. Access to mass transit includes the convenience of railway stations, Shinkansen express rail stations, and airports, and this access influences daily commuting as well as long-distance travel.

Hypothetical model and measures Model

The determinants of KIBS’ locations included the location determinants proposed by previous studies and the workplace selection criteria of enterprises, with which a hypothetical model was developed assuming the structure of the location mechanism. Among these variables, the concept of “regional image” was difficult to directly quantify, and therefore, it was indirectly measured using indicators of regional amenities regarding food, nightlife, entertainment, sports, shopping, and public parks along with measures of social tolerance (Figure ).

Measures

To develop the location model, the following measures were selected to operationalize the determinants . Table  lists the measures.

List of measures

“Important factors to workplace selection” (based on REWP)	Possible determinants of KIBS locations	Indicators	Data sources
Regional image	Social environment (tolerance)	Proportion of the literary and artistic population in the total working population	I
		Proportion of the in-migrant population in the total residential population	II
	Food amenities	Density of restaurant employees	I
	Nightlife amenities	Density of pub, beer hall, bar, and cabaret employees	I
	Entertainment amenities	Density of theater and other entertainment venue employees	I
	Sports amenities	Density of employees of sports facilities	I
	Shopping amenities	Density of total sales floors in the region	III
	Public park amenities	Average distance to the nearest parks (reciprocal)	IV
Access to other enterprises	Urban economies	Density of business enterprises at the municipal level	I
Employee transportation	Convenience of commuting	Average commute distance of workers (reciprocal)	V
Access to everyday resources	Convenience of everyday resources	Average distance to the nearest healthcare facility (reciprocal)	IV
Access to mass transit	Access to central station	Distance from the municipality to the Tokyo central station (reciprocal)	VI
	Social environment (talent)	Proportion of university graduates in the resident population	VII
	KIBS locations	Proportion of KIBS employees in the total workforce of the population	I

²List of data sources: I: 2009 EEC; II: 2009 Basic Resident Register; III: 2007 Commercial Statistics; IV: 2008 Housing and Land Survey; V: Fifth Tokyo Person Trip Survey; VI: 2000 National Census; VII: Distance in a straight line from the geographical centroid of the municipality to the Tokyo central station (measured with Arc GIS).

In the following sentences, details of the selected 14 measures that are listed in Table  are explained with respect to each of the seven elements in the hypothetical model (Figure ):

1. Regional image: social environment(tolerance)/urban amenities
   • a.Social tolerance was measured by 2 variables: (ii) proportion of the literary and artistic population in the total working population and (ii) proportion of the in-migrant population in the total residential population.
   • b.The extent of food amenities was measured as the density of restaurant employees.
   • c.The extent of nightlife amenities was measured as the density of pub, beer hall, bar, and cabaret employees.
   • d.The extent of entertainment amenities was measured as the density of theater and other entertainment venue employees.
   • e.The extent of sports amenities was measured as the density of employees of sports facilities.
   • f.The extent of shopping amenities was measured as the density of total sales floors in the region.
   • g.The extent of public park amenities was measured as the average of the distances from residences to the nearest parks.

Access to other enterprises

• Access to business partners located nearby was measured as the density of business enterprises at the municipal level.

Employee transportation

• This variable was measured by employees’ average commute distance using data drawn from the Fifth Tokyo Person Trip Survey.

Access to everyday resources

• Yamamura and Goto reported that access to everyday facilities was strongly inter-correlated among types of facilities and that overall access to them might best be measured by a single score on access to the nearest healthcare facility. Therefore, access to everyday resources was measured as the distance to the nearest healthcare facility.

Access to mass transit

• Access to distant business contacts was measured as the distance from the geographical centroid of the municipality to the Tokyo central station where many corporate headquarters are located.

Social environment (talent)

• A measure of the geographical distribution of talented people was included in the model measured as the proportion of university graduates in the resident population.

Location of knowledge industries

• Location of knowledge industries was indicated as the proportion of KIBS employees in the total workforce of the population.

Data reduction

Table  presents the correlations between the independent variables and the KIBS locations. Table  presents the correlations between the independent variables. The proportion of the literary and artistic population was not significantly correlated with geographical distribution of talented people, KIBS location, or the other independent variables. Therefore, that measure of social tolerance was determined to have no direct or indirect effect on the KIBS locations and it was dropped from the model.

Correlations between the independent variables and the KIBS’ locations

	Correlation coefficients with KIBS locations
Tolerance (writers, artists)	−0.003
Tolerance (in-migrants)	0.765
Urban economies	0.756
Convenience of commuting	−0.544
Convenience of everyday resources	0.693
Access to central station	0.645
Talented people	0.608
Food amenities	0.828
Nightlife amenities	0.766
Entertainment amenities	0.715
Sports amenities	0.846
Shopping amenities	0.715
Public park amenities	0.381

Correlations between the independent variables

	Public park amenities	Shopping amenities	Sports amenities	Entertainment amenities	Nightlife amenities	Food amenities	Talented people	Access to central station	Convenience of everyday resources	Convenience of commuting	Urban economies	Tolerance (in-migrants)	Tolerance (writers, artists)
Tolerance (writers, artists)	0.064	−0.053	−0.047	−0.015	−0.033	−0.030	−0.016	−0.042	−0.073	−0.023	−0.042	−0.108
Tolerance (in-migrants)	0.483	0.616	0.801	0.554	0.662	0.707	0.661	0.639	0.730	−0.589	0.675
Urban economies	0.320	0.757	0.843	0.765	0.888	0.952	0.348	0.932	0.608	−0.422
Convenience of commuting	−0.342	−0.428	−0.526	−0.350	−0.429	−0.464	−0.405	−0.376	−0.487
Convenience of everyday resources	0.560	0.568	0.755	0.506	0.641	0.679	0.567	0.574
Access to central station	0.313	0.589	0.753	0.584	0.806	0.863	0.312
Talented people	0.427	0.343	0.557	0.327	0.354	0.390
Food amenities	0.393	0.824	0.902	0.798	0.958
Nightlife amenities	0.405	0.759	0.862	0.711
Entertainment amenities	0.236	0.671	0.755
Sports amenities	0.453	0.791
Shopping amenities	0.345
Public park amenities

²⁰⁰P < .01.

Grey shadow: VIF > 10.

Multicollinearity results from strong correlations between independent variables, which degrades the reliability of model estimates. One index used to identify multicollinearity is the variance inflation factor (VIF). When the VIF is 10 or larger, there is a strong probability of multicollinearity.[Image Omitted. See PDF]

Substituting VIF = 10 into Equation 8 (above), R = 0.948. Therefore, in path analyses, model combinations of variables whose correlation coefficients are larger than 0.948 should not be included. Table  shows that correlation coefficients exceeding 0.948 were between food amenities and urban economies (0.952) and between food amenities and nightlife amenities (0.958). To eliminate the risk of multicollinearity, 1 of the 3 was chosen to measure all 3 of them. Because they all focused on urban elements that are typical in highly urbanized spaces, food amenities, which highly correlated with urban economies, nightlife amenities, and KIBS location, were chosen to measure all 3 and renamed “urbanity.”

Results of the path analysis

For the path analysis, the reduced hypothetical location model (Figure  above) described above was analyzed using structural equation modeling software (SPSS-Amos). Figure  illustrates the results. The coefficient of determination of KIBS location was 0.78, and the chi-squared test result was statistically significant (P < .001). However, the stability of the estimated coefficients was further assessed by repeatedly estimating the model in random samples . The path coefficients of these estimations were generally consistent with those of the original model , and the results of the initial path analysis were considered sufficiently stable.

Figure  illustrates the results on the KIBS’ locations as follows.

1. Geographical distribution of talented people and urbanity (food amenities, nightlife amenities, and urban economies) had strong direct influences on KIBS’ locations. The influence of the geographical distribution of talented people was consistent with Florida's results. The influence of urbanity was even stronger than that of the geographical distribution of talented people.
2. Social tolerance and sports amenities positively influenced the geographical distribution of talented people. The influence of social tolerance is consistent with Florida's results. The influence of sports amenities was even stronger than that of social tolerance.
3. The relationship of access to mass transit to KIBS’ location was negative, meaning that access to central Tokyo was not a factor that was important to location.

Comparing Figure  to Figure , which shows the results on non-KIBS’ locations, the relative features of the location mechanism of KIBS are understood as follows.

1. The influence of the geographical distribution of talented people was weaker on KIBS location than on non-KIBS location, suggesting that the geographical distribution of talented people was not particularly attractive to KIBS enterprises, at least at the intra-metropolitan scale in the TMA .
2. The influence of urbanity (food amenities, nightlife amenities, and urban economies) on KIBS location was observed as stronger than its effect on non-KIBS location. Therefore, a societal context that includes many amenities and urbanization is a major location determinant for KIBS enterprises.

However, urbanity was a multi-dimensional concept that used one variable as a proxy measure of 3 factors, which was a methodological choice made to avoid the potential problems of multicollinearity. Consequently, this result does not specifically identify which factors, among food amenities, nightlife amenities, and urban economies, had the strongest influences on KIBS location. To improve the precision of the results, a questionnaire survey of KIBS enterprises was conducted, and numerous factors, including the 3 dimensions of urbanism, were analyzed regarding the choice of location.

Questionnaire survey Survey sample

To choose a sample of KIBS enterprises for the survey, all of the 20 municipalities with an LQ larger than 1.0 (see Section  above) were listed. Next, using a random numbers table, 200 KIBS enterprises in those municipalities (within a total of 36 648) were chosen from the Town-Page (classified telephone directory) Internet database. These 200 enterprises were contacted by telephone and asked to participate in the survey. Ninety-four enterprises agreed to participate (47%). The questionnaire was sent to these enterprises by post, and the answer sheets were returned in the provided stamped self-addressed envelopes. Of the 94 mailed questionnaires, 56 valid responses were returned (60% of the mailed questionnaires). Table  provides detailed information about the questionnaire.

Details of the questionnaire survey

Method of sample selection and request for the survey	First, 200 KIBS enterprises were randomly sampled (using a random numbers table) from the NTT Town-Page Internet database that are located in the 20 municipalities with an LQ larger than 1.0. Next, these 200 enterprises were contacted by telephone and asked to participate in the survey. Ninety-four enterprises agreed to participate.
Distribution period	June 15–16, 2012
Reply period	June 17–July 10, 2012
Reply method	Mail (with a return-mail envelope enclosed with request documents)
Survey items	Important factors to enterprises’ locations Needs of meetings outside of meeting rooms Desirable regional image Desirable work environment for employees
Number of valid responses	56 (60% of the mailed questionnaires)

Most of the respondents (86%) were in top management and had knowledge on and decision-making power regarding the locations of their enterprises, which are the qualifications of respondents requested in the questionnaire's cover letter.

Reasons to choose a location

Figure  illustrates the variables that are important to the respondents when determining their enterprises’ locations. Some 86% of the respondents reported that proximity to clients was important, followed by proximity to partner enterprises (66%) and proximity to outsourcing, subcontractors, and suppliers (66%). These results indicate that the direct benefits of proximity to other enterprises were an important location determinant. In addition, the respondents reported that restaurant quality suitable to business meetings (64%), receptions and dinners with partner enterprises (61%), and employees’ lunches (75%) was important. Thus, food amenities for business and individual purposes were important location determinants. The regional image was another important location determinant because many respondents emphasized regional image from the business perspective (89%) and from the employee perspective (84%).

Among the factors that were highly important to the respondents, the proximity to business enterprises implies that KIBS are attracted by proximity benefits, but this study did not consider them in depth. Proximity benefits are economic gains from proximity to other enterprises, and their characteristics and causes have been widely studied. Therefore, this study considered other factors based on the survey results.

Public meetings

The respondents’ emphasis on restaurant quality suitable to receptions and dinners with partner enterprises suggest a strong interest in meeting in public at restaurants near their workplaces. To confirm this, the survey results were examined on the question, “Do you frequently have meetings at nearby restaurants, cafes, or tearooms?” Approximately 84% (n = 47) responded “yes.” In addition, the responses to “Why do you have meetings outside of meeting rooms?” indicated that 57% of the respondents were using restaurants in response to a lack of office meeting space, meaning that they were using restaurants as extensions of their workplaces (Figure ).

Work environment

Strong emphases on restaurant quality suitable to employees’ lunches and regional image from the employee perspective suggest that providing adequate work conditions for employees was important for location choice. To confirm this, the responses to “Do you consider employees’ work satisfaction when you determine workplace location?” were examined. The vast majority responded “yes” (91%, n = 51). Furthermore, the responses to “What factors do you think might improve the work satisfaction of your employees?” indicated that respondents responded “good regional image (84%)” and “quality of nearby restaurants (57%)” (Figure ), which is consistent with the survey results shown in Figure  above.

Regional image

Regional image was an important location determinant understood as a key factor for improving the corporate image held by partner enterprises and employees’ work satisfaction. The responses to “What is a desirable regional image for your business location?” allowed for multiple responses from a 5-option multiple choice question (Figure ) , and the most commonly chosen option was “urban.”

Conclusions

In this study, KIBS were more densely agglomerated than non-KIBS. There were several densely agglomerated KIBS clusters in central Tokyo and in some of the inner suburban business districts. The EG of the KIBS enterprises was 0.023562, compared to 0.000148 among the non-KIBS. This result quantitatively confirms that the KIBS in the study were more likely than the non-KIBS to geographically concentrate.

The most influential factor on KIBS’ location was urbanity (food amenities, nightlife amenities, and urban economies), and the effect of urbanity on KIBS’ location was stronger than it was on non-KIBS’ location. Thus, a social context with high urbanism was an important absolute and relative location determinant for KIBS. The second most influential determinant was the geographical distribution of talented people, which was positively influenced by social tolerance and sports amenities and negatively influenced by urbanity. However, the influence of the geographical distribution of talented people on KIBS’ location was weaker than its effect on non-KIBS’ location. Therefore, although the geographical distribution of talented people was attractive to KIBS, it was not characteristic of KIBS’ location choices. Additionally, the coefficient of the relationship between access to mass transit and KIBS’ location was negative, indicating that access to central Tokyo was not important.

After analyzing the aggregate data, a questionnaire survey was conducted on KIBS enterprises (through their executives) to investigate the proxy measure, urbanity, which was found to be the most influential location determinant. The survey data found the following 4 major results.

1. Approximately 86% of KIBS considered proximity (which is expected to provide benefits through mutual interaction) to be an important factor for location determination.
2. Local restaurants, cafes, and tearooms were reported as often used as workplace extensions because of a lack of space for meetings in the offices.
3. Employees’ work satisfaction was a high priority when KIBS determined workplace locations, and a good regional image and the quality of nearby restaurants were major contributing factors.
4. A sophisticated and urban regional image was preferred by the KIBS respondents because it was understood as a positive influence on the corporate image.

The results of the questionnaire survey imply that, in highly urbanized areas, the dense distribution of business enterprises, food and nightlife amenities, and an urbane regional image create a positive urban environment evaluation by KIBS enterprises and attract them through direct and indirect causal relationships (Figure ).

Disclosure

The authors have no conflict of interest to declare.

Notes:

¹⁰⁰³According to Yada, with increasingly fierce competition and the development and diffusion of microelectronics, which accelerated innovation, knowledge production gained an important supportive role and the “knowledge industry” and “knowledge occupations” rapidly grew (p. 154). Knowledge intensification is advancing in consumer services as society becomes more complex and consumer needs diversify (p. 153).

¹⁰⁰⁴See p. 40 of Miles.

¹⁰⁰⁵“High-tech industries” are significantly overlapped with the industrial classifications defined as KIBS in this study. For example, Saxenian's main research subjects were software development and Internet firms, which were KIBS.

¹⁰⁰⁶The TMA was defined as the survey area of the Fifth Tokyo Person Trip Survey, comprising Tokyo, Chiba, Saitama, and Kanagawa Prefectures and the southern region of Ibaraki Prefecture.

¹⁰⁰⁷The employee size criterion used to exclude huge enterprises was 100 employees. That limit was chosen because, according to the Small and Medium Enterprises Basic Act, service business enterprises are defined as “small and medium sized” when they do not exceed 100 employees and as “large” when there are more than 100 employees. The EEC data also classify workplaces by the 100-employee cut-off.

¹⁰⁰⁸Previous studies have found that, unlike other knowledge-oriented industries, the major determinant of location among Research and Development Institutions (JSIC-71) was proximity to governmental research institutions. Because of their unique location characteristics, it was inappropriate to include Research and Development Institutions with the other non-KIBS industries, and they were not included in the analysis.

¹⁰⁰⁹Appendix Table A1 lists the selected 2-digit industrial classifications.

¹⁰¹⁰Appendix Table A2 lists the excluded three-digit industrial classifications.

¹⁰¹¹“Service industries” were identified through the following JSIC classifications: 37–49, 68–70, 72–80, 82–85, 88–92, and 95.

¹⁰¹²LQ(i) in region A = (Number of employment in an industry “I” of region “A”/Number of employment in an industry “I” in the TMA)/(Nighttime population of region “A”/Total nighttime population of TMA).

¹⁰¹³For detailed explanations of the Ellison-Glaeser index, see p. 899 in Ellison and Glaeser and Nakamura.

¹⁰¹⁴Two factors were excluded because they primarily influence location at the neighborhood or smaller scale: (i) visibility from street and (ii) distance from the nearest train station.

¹⁰¹⁵When selecting some measures, we referred to the following previous studies: (i) regarding the geographical distribution of talented people (Shearmur et al.), (ii) regarding access to other enterprises (Otsuka et al.), (iii) regarding access to everyday resources (Yamamura and Goto), and (iv) regarding the proportion of the literary and artistic population (Florida).

¹⁰¹⁶Sampling was conducted as follows: (i) a total of 121 cases were chosen using a table of random digits from the 242 cases of the original dataset, (ii) path analysis was performed using the 121-case dataset, (iii) repeated sampling and path analyses were performed 10 times, and (iv) deviations in the coefficients of the 10 models from the path coefficients of the original model were checked.

¹⁰¹⁷Appendix Table A3 shows the results of the random re-sampling.

¹⁰¹⁸The results found that the effect of geographic distribution of talented people on KIBS’ location was weaker than its influence on non-KIBS’ locations. These results seem inconsistent with Florida, who argued that high-tech industries are attracted to locations with high proportions of highly educated people. However, the inconsistency might relate to the difference in scale between the 2 studies (eg, this study found the determinants in the interior scale of the metropolitan area, whereas Florida found the determinants through an inter-city scale). This study used municipal-level data to measure the proportion of university graduates in the residential population as a proxy for the geographical distribution of talented people; therefore, the causal relationships tested in the model were at the municipal level. Thus, in addition to the relationships found by this study, the dynamics at higher levels might be possible in which highly educated workers attract KIBS across municipal boundaries.

¹⁰¹⁹We refer to Goto concerning 5 factors of regional images.

Appendix A1 List of selected industrial classifications

39	Information Services
40	Internet Based Services
41	Video Picture, Sound Information, Character Information Production and Distribution
70	Goods Rental and Leasing
72	Miscellaneous Business Services
73	Advertising
74	Miscellaneous Business Services
88	Waste Disposal Business
90	Machine, etc. Repair Services, Except Otherwise Classified
91	Employment and Worker Dispatching Services
92	Miscellaneous Business Services
93	Political, Business and Cultural Organizations

A2 List of excluded industrial classifications

705	Sports and Hobby Goods Rental
709	Miscellaneous Goods Rental and Leasing
741	Veterinary Services
746	Photographic Studios
909	Miscellaneous Repair Services

A3 Result of the random re-sampling

	Original model	Path coefficients of resample models (n = 121, 10 times random sampling)
	Path coefficients	Ave.	SDV
Tolerance > talented people	0.45	0.45	0.07
Sports amenities > talented people	0.49	0.48	0.09
Entertainment amenities > talented people	0.03	0.06	0.07
Urbanity > talented people	−0.44	−0.43	0.09
Shopping amenities > talented people	−0.07	−0.09	0.07
Public park amenities > talented people	0.11	0.11	0.07
Tolerance > KIBS	0.15	0.13	0.04
Sports amenity > KIBS	0.09	0.11	0.14
Entertainment amenities > KIBS	0.05	0.08	0.09
Urbanity > KIBS	0.84	0.82	0.07
Shopping amenities > KIBS	−0.06	−0.09	0.06
Puclic park amenities > KIBS	−0.10	−0.08	0.04
Convenience of commuting > KIBS	−0.09	−0.09	0.02
Convenience of everyday resources > KIBS	0.04	0.06	0.08
Access to central station > KIBS	−0.29	−0.29	0.07
Talented people > KIBS	0.31	0.28	0.03
Tolerance > non-KIBS	0.06	0.07	0.07
Sports amenities > non-KIBS	0.01	0.00	0.13
Entertainment amenities > non-KIBS	−0.25	−0.24	0.08
Urbanity > non-KIBS	0.44	0.38	0.17
Shopping amenities > non-KIBS	−0.08	−0.05	0.11
Public park amenities > non-KIBS	0.04	0.02	0.06
Convenience of commuting > non-KIBS	0.32	0.31	0.04
Convenience of everyday resources > non-KIBS	−0.01	−0.01	0.08
Access to central station > non-KIBS	−0.36	−0.35	0.09
Talented people > non-KIBS	0.39	0.40	0.08