Customer satisfaction with Tokyo Metropolitan Government ward oce counter services: conrmatory factor analysis of question grouping and principal component regression analysis

This study examines the determinants of service satisfaction among users of a ward oce in Tokyo using a two-part questionnaire. The questionnaire comprised three main categories of multiple-choice questions: A) facility equipment (physical elements in the government oce), B) staff responses, and C) service delivery (e.g., promptness of services, whether visitors completed their scheduled errands). In addition, three groups of questions related to the personalities of the users of each facility were investigated. During a one-day survey period, responses were collected from 400 women, aged 30–59 years, who had visited a ward oce in the Tokyo Metropolitan Area over a one-month period. This age segment was studied because it is the major segment of the panel of the Internet research rm used. First, factor analysis was used to check the appropriateness of the grouping of questions, and it was conrmed that the three groups were appropriate based on eigenvalues and scree plots. Then, to examine the determinants of counter service customer satisfaction, principal component analysis and multiple regression analysis were conducted for each question category. The regression analysis using the three main question groups and three other respondents’ personality-related question groups indicated that group C (service delivery quality) had the strongest inuence on the dependent variable, followed by group B and group A. The adjusted R 2 value was .70. This result is consistent with the results of the author’s surveys of government oces in urban areas conducted for Higashihiroshima City Hall and of ward oces in Osaka.


Introduction
In the world of public administration, the concept of governance has attracted considerable attention in recent decades. The rst issue of the journal Governance: An International Journal of Policy, Administration, and Institutions (1988) contained a number of articles on the changing relationship between politics and public administration, which has been the subject of ongoing debate in American public administration research, such as in Campbell and Peters (1988). Subsequent research on public administration has been divided into two major categories: (1) the use of governance as an abstract concept that represents a change in the way the entire administration is governed, and (2) the use of governance as a concrete concept in the same way as corporate governance, in the sense of compliance with laws and regulations and proper operational management of organizations. Example of the former are Compton, et.al. (2021) and Dickson (2016), and examples of the latter are Brunet and Aubry (2016) and Almqvist et al. (2013). When the concept of governance is used in the private sector, it is often used in a speci c meaning related to the latter scal accountability. While governance in the corporate sector means enacting rules and strengthening the authority of shareholders and boards of directors, discussions of governance in the government sector often mean a retreat or reduction in the oversight and role of government. Along with the spread of the concept of governance, a major area of growth in schools), while the lower-tiered local governments (municipalities equivalent to cities, towns and villages in other prefectures), the special wards, provide familiar administrative services such as resident registration, garbage collection, and operation of elementary and junior high schools.
In this study, in addition to the three groups of questions used in the author's previous studies (Moteki, 2021), A (facility equipment), B (staff response), and C (service delivery), three additional groups of questions related to the personalities and thoughts of the facility users were added. B (staff response) is a question group related to the explanation of the counter staff and their treatment of customers. C (service delivery) is a question group about the status of service delivery related to the completion of errands at the ward o ce, the speed of those errands, and the waiting time. Research on customer satisfaction in both the private and government sectors is no longer limited to the provision of services and their quality, but also includes the attributes and psychological state of service providers and purchasers (users) as a factor in customer satisfaction (e.g., Nazi, 2010). Yu et al. (2011) quantitatively examined the effect of service recipient attributes such as household income on satisfaction with the service. Vermeeren et al. (2011) used the metaphor of a "satisfaction mirror," meaning that employee satisfaction in uences customer satisfaction, and mentioned that educating employees about the signi cance of a company's business programs is important as an instrument of internal marketing. They examine data on work environment characteristics, job satisfaction, and customer satisfaction of frontline employees in 35 Dutch municipalities in relation to data on customer satisfaction. Smith (2020) examined the effect of service purchasers' personality attributes on satisfaction and brand loyalty for cell phone services. Rha (2012) conducted a questionnaire survey in South Korea, in the eld of social work, childcare, and healthcare services, using a 5-point questionnaire based on SERVPERF. In addition to service quality and service delivery, "relationship quality," de ned as "the depth and climate of the relationship between suppliers that participate in the public service delivery" and based on Johnson (1999), was examined, and it was argued that this variable could in uence customer satisfaction as a mediating variable.
Professor Yu Noda (Noda, 2013; has published extensive research on Japanese citizen satisfaction with public administration in English. The author's research shares Professor Noda's awareness of the problem, but is on a more speci c level and differs in terms of methodology. Speci cally, the author's study employed a questionnaire survey to examine the determinants of customer satisfaction with the counter services provided at local government o ces, using a multi-group questionnaire developed from SERVQUAL (to be discussed later) to quantitatively and concretely examine the factors that contribute to satisfaction. It applies the multi-group question survey method with a 7point scale, which is widely used in the eld of business administration, to public administration, and conducts empirical research at the level of speci c government facilities. Everitt (2005) also mentions performing principal component analysis on multiple question items belonging to several categories before multiple regression analysis. This avoids the risk of multicollinearity between each question item. By surveying speci c users of Japanese government o ces about their experiences using a questionnaire consisting of three groups of questions (a simpli ed version of SERVQUAL using ve groups of questions), I would like to show that the service quality evaluation methods widely applied in the private sector can also be used in the public sector. This is accomplished by examining the results of con rmatory factor analysis and regression analysis, especially coe cients of determination and standardized partial regression coe cients.

Literature Review
The study of citizen satisfaction with services in public administration has been in uenced by similar studies in the private sector, which developed earlier. Several important studies on customer satisfaction in public administration have been reviewed previously by Moteki (2020). In the following review, I will focus on particularly important literature, as well as new literature not previously covered. Wagenheim and Reurink (1991) explained the difference between internal and external customers (employees in other departments; citizens), and discussed that the satisfaction of the service needs of these two groups leads to organizational e ciency and effectiveness.  Huque & Hayllar, 1999). This study referred to the framework used by Mansor and Razali (2010) in Malaysia. The present study uses the questionnaire in Moteki (2020), in which speci c question items were grouped into three categories, originally developed by Mansor and Razali (2010). Mansor and Razali (2010) provide groupings and a broad description of each, but do not provide details within the paper of the speci c questions for each group or how the survey was conducted.
Before the advent of SERVQUAL, Oliver's prominent study, focusing on the discrepancy between expectations and subsequent perceptions of performance as a determinant of customer satisfaction (Oliver, 1980;Oliver & Winer, 1987) and conclude that it is appropriate to include citizen trust, rather than user satisfaction, as a soft indicator for this purpose. Using a general mass survey of Norwegian citizens conducted in 2001, Christensen and Laegreid (2005) clari ed several points concerning trust in government, including that citizens who were satis ed with public services tended to have trust in public institutions. Heintzman and Marson (2005) similarly assume that citizen trust and con dence in public institutions are generated by citizen/client service satisfaction. They argue that service satisfaction, employee satisfaction, and commitment are mutually in uential, as shown in Figure 1, which shows the public-sector service value chain. Korunka et al. (2007), using their own scale of customer orientation through long-term questionnaire surveys in Australia and the United States, showed that customer orientation among employees is essential for customer satisfaction. Mouwen (2015) used data for 2010-2011 in the Netherlands on public transportation, including buses, subways, trams, and ferries, to examine whether user satisfaction is affected by the characteristics of the transportation mode and/or the attributes of the users.

Methods
To examine the determinants of satisfaction with administrative counter service, we prepared a questionnaire that can be divided into three groups with respect to the explanatory variables. The questionnaire comprised three main categories: physical aspects of facility equipment, counter staff responses, conditions of service delivery. The classi cation was based on the three explanatory categories used by Mansor and Razali (2010). There were ten survey questions and two screening questions concerning respondents' conformity to the survey conditions. Table 1 illustrates the main question items regarding counter service satisfaction in government o ces, classi ed into the three concept groups. The questionnaire also asked for responses to items on individual personality and characteristics-(M) thoughts about management aspects of public administration, (I) international tendency, and (F) authoritarian tendency-and posed a nal open-ended question regarding how to improve ward o ces. The name of question group F was drawn from the scale of authoritarianism presented by Adorno et.al. (1950), which is called the F Scale. These three additional sets of questions were used for exploratory analysis, to inform more detailed future studies. It has been pointed out that there is a connection between the NPM management orientation and authoritarianism and populism (Massey, 2019;Harun et al., 2021). In other words, the NPM management orientation is consistent with the authoritarian orientation among people and the populist tendencies in democracy, according to Prior to the screening questions regarding respondents' attributes, the following information was clearly displayed on the screen: (1) the author was responsible for carrying out the survey and was willing to answer questions about it (including questions about the results of the survey, to be published in academic journals), through a web form, and (2) the survey participation was voluntary and was not related to the city government; respondents would not be judged unfavorably based on their survey responses, and an explanation of the privacy protection policies and terms was provided at the beginning of this online survey.

Results
The survey was conducted on a weekday (December 1, 2021). There were a total of 400 respondents: 162 (40.5%) were in their 30s, 129 (32.3%) in their 40s, and 109 (27.3%) in their 50s (because of rounding, the total does not add up to 100.0%). The collection time was just after the typical lunch break in Japan, from 13:03 to 14:20, a little over an hour, until the planned number of respondents was gathered. Table 3 shows the respondents' reasons for visiting the ward o ce (they were asked to choose only one). This shows that the largest percentage of residents visited the ward o ce for matters related to family registration or resident registration.  (2016) conducted a factor analysis for each question and found that it was appropriate to divide the questionnaire into three groups. Using question items A-C, the maximum likelihood method was used for factoring, and the Varimax method was used for factor rotation. The eigenvalues were, in order from the rst factor, 7.732, 1.992, 1.229, and 0.796. Up to the third factor, the eigenvalue is greater than or equal to 1 (the Kaiser-Guttman criterion). The number of factors was determined to be 3 according to both the Kaiser-Guttman criterion and the scree plot criterion. The scree plot of Figure 2 shows that the third factor is valid. Looking at the results of the factor analysis in Table 4, the three classi cations based on the factor scores of each question item are basically consistent with the three classi cations assumed by the author. For Q4_6, regarding infection control measures such as sanitizing solutions and acrylic plates for COVID-19, factor scores calculated from factor analysis exceeded 0.4 for the rst and second factors. This means that question Q4_6 is an item that relates to both Factor 1 and Factor 2 (Group A and B in the author's questionnaire classi cation). What this means is that infection control measures such as sanitizing solutions and acrylic panels for COVID-19 involve questions about hardware as well as the nature of the responses by staff; infectious diseases can be passed from person to person, and staff members sometimes talk to each other about sanitizing. In addition, both government agencies and private establishments often have problems regarding disinfection and saliva spray, and such recent circumstances during the pandemic may be re ected in this. As shown in Table 5, the correlation coe cient between the second and third factors is .060, but this and the other correlation coe cients between different factors in this table are not signi cant at the 5% level.
From this, we can assume that each factor is independent. The results of the factor analysis so far are similar to those of Talib and Shukor (2016), indicating that the grouping of three groups is appropriate in terms of content homogeneity across questions, less than SERVQUAL's ve groups. Next, a principal component regression analysis was performed using the three groups of questions (main items), whose grouping had been con rmed as appropriate by the factor analysis.  Table 6 shows the correlation coe cients between each question item in groups A to C and ZY1. The following principal components analysis for each group uses the marked question items, with a correlation coe cient of 0.4 or higher. However, for groups I, M, and A, which are additional questions, all question items are used as they are for the principal component analysis, because the structure and content of the questions for each of the three scales have not yet been fully explored and will be used for exploratory purposes only (the results of the correlation analysis with ZY1 are shown in Table 7).      The variance in ation factor (VIF) of ZB1 was slightly greater than 2.00. This means that there is an overlap between the set explanatory variables (i.e., the risk of multicollinearity) in group B. Next, multiple regression analysis (Model 2) was conducted by adding the principal component scores of the variables related to the personalities and thoughts of the users (M, I, and F) in addition to the explanatory variables of the three groups (A-C). Table 9 shows the results of the regression analysis in Model 2. The adjusted R-squared value of the model's coe cient of determination is .71, which is higher than that of Model 1. At the 1% level, ZC1 and ZA1 had positive effects on the explained variable ZY1, in that order. The VIF for VB1 was slightly over 2.00; therefore, some concerns remain regarding independence among the selected explanatory variables. The principal component score variables calculated from the I and F groups (ZI1, ZI2, and ZF2) had coe cients that were signi cant at the 1% level in this model. The rst principal component of group I (international orientation) had a positive coe cient, the second principal component had a negative coe cient, and the standardized partial regression coe cient (B) was small but statistically signi cant. Because the adjusted R-squared value of the model's coe cient of determination is higher than that of Model 1, we can say that Model 2, with the respondent's personality and thoughts variables, has more explanatory power than Model 1.

Discussion
The  (Table 3) demonstrated that ZC1 from group C (service delivery) was the most in uential (β = .50), followed by ZB1 (the rst component of customer service provided by the counter staff) and ZA1 (the rst component of o ce hardware), with an adjusted R2 value of .70. For explanatory variables A-C, the most in uential variable was ZC1. The regression coe cients for each variable indicated that the most important variable was service delivery quality (group C). This result was inconsistent with the results of Mansor and Razali (2010) in Malaysia, which revealed that the human factor was more important than administrative services themselves. Our study results of the 2020 survey at the Kurose branch o ce of Higashihiroshima City government similarly showed that the human factor was the most in uential, and the results differed from the results of this survey of ward o ce users in Tokyo and Osaka. This difference would indicate the regional factors of urban and rural areas and the different authorities that government o ces have. In this pandemic era of widespread coronavirus infections, both private and public organizations are increasingly required to improve their service delivery methods and service quality (Sheth, 2020). In future research, I would like to examine how the changes brought about by COVID 19, including the infection prevention measures in the facilities discussed in this paper, affect the factors that contribute to customer satisfaction, or the factors that improve service quality, and the psychological conditions and behavior of users and employees in the facilities.
In this study, we additionally asked three groups of questions (I, M, and F) related to the respondents' personality and thoughts, which were used in the multivariate analysis. For the rst time, a con rmatory factor analysis was conducted to examine the appropriateness of the grouping of explanatory variables.
In light of criteria consisting of eigenvalues and the Kaiser-Guttman criterion, three categories were found to be appropriate, as expected. Thus, we wrapped up each of them as a compound variable through principal component analysis and analyzed their correlation with the explanatory variable ZY1. Next, we conducted a multiple regression analysis of customer satisfaction, taking into account the personality and thoughts of the facility users, and were able to construct a model with more explanatory power in terms of the coe cient of determination (Model 2: Adjusted R 2 .72) than when analyzing only the variables of the administrative side involved in the service (Model 1). As mentioned in the introduction and the related research section, factors that contribute to customer satisfaction are not only the quality of the services provided, but also the circumstances and attributes of the users themselves and the service providers or the counter staff. Looking at the results of the regression analysis as described here, especially the statistical signi cance of its coe cient of determination and standardized partial regression coe cients, it can be understood that it is possible to use the service quality evaluation method widely applied in the private sector, such as that attempted in this study, in the public sector.
Differences between public and private organizations include the strength of regulation by law and the ambiguity of goals (Rainey, 2010). In the future, we would like to consider and explore more appropriate This study set three main categories: hardware, counter o cer responses, and service delivery quality (groups A-C). Additional aspects related to other customer satisfaction factors may be needed for future research, to develop a more explanatory model. Items with less than 0.4 correlation coe cients with the composite variables Y group were not subjected to principal component analysis. To explore models and questionnaires that can better grasp customer satisfaction at government o ces, we need to conduct a follow-up survey with renewed and re ned question items. By comparing the results of these surveys with those of such future studies, the determinants of customer satisfaction could then be explored in detail.
The author conducted two on-site surveys in government buildings and three Internet surveys on customer satisfaction in public administration in Japan. On-site and Internet surveys each have advantages and disadvantages. The advantages of the on-site survey are that (1) the researcher can answer respondents questions about the questionnaire directly and (2) the researcher can check the actual situation of the government o ce during the survey period and identify unexpected events such as angry users. Internet surveys also have their merits, the most important being the ability to eliminate the time and effort involved in processing data, including data entry and tabulation. Another advantage is that, as shown in this paper, it is relatively easy to conduct surveys that include matters related to the respondent's personality and attributes, which are di cult to obtain in face-to-face surveys. Although we were unable to investigate this in detail in this analysis, the author would like to examine whether the mode of survey (on-site or Internet) has any effect on the results of the satisfaction survey.
The author conducted a quantitative survey in the eld of Japanese local governments using a questionnaire consisting of three original groups of questions, which are simpli ed versions of the ve groups of SERVQUAL. In this regard, there are ways to focus on the exceptional case of angry customers, as in Bougie et al. (2003). These people are called "claim" or "claimer" in Japanese language, with a negative meaning, and the problem of trouble between citizens and counter staff developing from customer complaints is becoming more and more prevalent at administrative o ces in Japan. The troubles described here seem to be caused by various reasons, such as (1) long waiting time, (2) unful lled wishes, (3) original distrust that users have for government agencies, (4) users' dissatisfaction with society and government agencies, and (5) users' bad health conditions. One customer's problem can often affect the satisfaction of other customers and the motivation of the administrative staff, because the presence of an angry user on the oor will inspire anger and anxiety in other users. Service recovery, meaning recovery or resolution from a failure in service delivery that caused customer outrage or dissatisfaction. has begun to be discussed in business administration as a way to deal with customer complaints, customer service failures, and errors (Johnston and Michel, 2008). For the present kind of study, dealing with customer complaints and service recovery in public administration, a qualitative study involving detailed investigation of individual cases and individual interviews with staff and citizens would be useful. In addition to the quantitative methods using questionnaires that we have focused on so far, we would thus like to add qualitative methods, such as interviews, to further work on the study of customer satisfaction in public administration in Japanese settings.

Declarations
Competing interests: The author declares no competing interests.  Factor loadings related to A Figure 5 Factor loadings related to B Figure 6 Factor loadings related to C Figure 7 Factor loadings related to M

Figure 8
Factor loadings related to I Figure 9 Factor loadings related to F