1 Introduction

Research on subjective well-being in China is still embryonic, relative to the large literature that has emerged in Western contexts (Chen and Davey 2008a, b). One promising line of inquiry is the application of the Personal Wellbeing Index (PWI) to Chinese samples. The PWI is a multi-item indicator of subjective well-being first developed in Australia as part of the Australian Unity Wellbeing Index (Cummins et al. 2003). As of 2005, the PWI was being used by over 100 researchers in 50 countries (International Wellbeing Group 2006). The psychometric properties of the PWI are well established in a series of studies with western samples (International Wellbeing Group 2006). The PWI has been found to have similar psychometric properties to its use in Western samples when applied in Hong Kong (Lau et al. 2005, 2008), Macau (Macau Inter-University Institute 2007), urban China (Chen and Davey 2009; Huang and Xing 2005; Smyth et al. 2009), rural China (Davey et al. 2009), to off-farm migrants in China (Nielsen et al. 2009) and the Tibetan plateau (Webb 2009).

While the results from these initial applications of the PWI appear promising, there have been several calls to undertake further empirical research on the application of the PWI to Chinese samples. For example, Chen and Davey (2009) stated, “the validation of the [PWI] for Chinese populations should be deemed as a work in progress, and ….[is] expected to undergo evolution in response to further empirical evidence”. Davey et al. (2009) have echoed these sentiments: “Although research [applying the PWI in China] has reached interesting findings, it is in an early stage of development compared to Western research. …. Further research is needed from a range of demographic groups to gain an overall understanding of Chinese citizens’ subjective quality of life. …. China is a country with considerable diversity in its land and people, and there are ample opportunities for further work.”

The theoretical underpinning for the PWI is the “Theory of Subjective Wellbeing Homeostasis” (Cummins 1998, 2000; Cummins and Nistico 2002; Cummins et al. 2002, 2009). This theory proposes that, under normal circumstances, subjective well-being is maintained within a limited positive range by neuro-psychological mechanisms analogous to the homeostatic management of body temperature. Cummins et al. (2009) list several characteristics which should be displayed if subjective wellbeing management is homeostatic. There must be a threshold value which is defended by the homeostatic mechanism and there must be evidence that, as this value is approached, the system works harder than normal to retain control. Once the threshold value is passed, there must be evidence that the homeostatic mechanism is defeated and is no longer controlling subjective wellbeing. For example, there is evidence that defeat of the homeostatic mechanism results in reduced subjective wellbeing and high depressive symptomatology (see Davey 2004). If the homeostatic mechanism is defeated, over time the system should attempt to regain control and, if it does regain control, subjective wellbeing should return to a stable set-point. However, as the homeostatic mechanism has a limited capacity to recover normal functioning, if the negative challenge is chronic and strong, recovery may not take place.

The objective of the homeostatic mechanism is to maintain subjective wellbeing within a narrow set-point range. Adverse environmental factors that have the potential to defeat the homeostatic mechanism “may have their origins either external to the person, in terms of life events, or within the person, such as in perception of pain” (Cummins et al. 2002, p. 13). Both internal and external buffers exist which can assist the homeostatic mechanism to prevent subjective wellbeing falling below the threshold value (Cummins et al. 2009). External buffers include wealth and personal relationships, which both serve to moderate the influence of potential stressors on subjective wellbeing. Internal buffers comprise adaption and habituation protective devices through which individuals use cognition to restructure reality, so as to minimize the impact of unavoidable negative experiences.

For western samples the mean PWI is 75 and the standard deviation 2.5, on a 0–100 scale distribution. The normative range has been found to be 70–80 points—a distance of two standard deviations either side of the mean (Cummins et al. 2003, 2004, 2009). These values are generally about 10 points lower in Chinese samples, which has been attributed to cultural bias (Lau et al. 2005; see also Chen and Davey 2008a, b). Webb (2009) found that the overall magnitude of life satisfaction of ethnic Tibetans was more consistent with Western than Chinese populations, but his sample was small (n = 102). One instance in which subjective well-being was found to fall below the normative range predicted by the “Theory of Subjective Wellbeing Homeostasis” was Algeria for which the PWI was 52.30 (SD = 21.10). This finding was explained in terms of the existence of adverse environmental factors, such as severe economic and political instability, which altered Algerians perception of reality and defeated the homeostatic mechanism (Tiliouine et al. 2006).

This paper responds to calls for research on subjective well-being using more diverse demographic groups in China through reporting the findings of a study that administered the PWI to a sample of Beijing taxi drivers in the lead-up to the Beijing Olympic Games in 2008. In previous research, Nielsen et al. (2009) argued that China’s off-farm migrants represent a particularly strong test of the “Theory of Subjective Wellbeing Homeostasis” because of the harsh environmental conditions which they are forced to endure. Nielsen et al. (2009) found that the subjective well-being of off-farm migrants was in the range predicted by the theory 62.56 (SD = 14.64). Examining the subjective well-being of Beijing’s taxi drivers, who work long hours for low wages, represents a similarly strong test for the “Theory of Subjective Wellbeing Homeostasis”. The Beijing Olympic Games represented a further external shock that intensified pressure on Beijing’s taxi drivers and potentially impacted on their homeostatic mechanism. If the PWI for Beijing taxi drivers is found to lie within the normative range predicted for Chinese societies under these circumstances, this would provide further evidence that the homeostatic mechanism is resilient in Asian contexts.

The specific aims of this research are as follows:

  1. (a)

    Report on the PWI in terms of its psychometric properties.

  2. (b)

    Assess whether Beijing taxi drivers are satisfied with their lives.

  3. (c)

    Examine whether the responses to the PWI from participants falls within the narrow range predicted by the “Theory of Subjective Wellbeing Homeostasis”.

2 Methodology

2.1 Instrument

The PWI was used in the present study to measure domain-level representation of subjective life satisfaction. The Chinese (Mandarin) version of the PWI, which is available on the International Wellbeing Group website, was used for this purpose. It has previously been used in several studies (see Huang and Xing 2005; Chen and Davey 2008b). The PWI used in the present study consisted of seven domains, measured on an 11-point end defined Likert scale, with numerical ratings ranging from 0 (extremely dissatisfied) to 10 (extremely satisfied). The seven domains were standard of living, personal health, achievement in life, personal relationships, personal safety, community-connectedness and future security. An eighth domain focused on religion and spirituality added to the PWI in November 2006 (International Wellbeing Group 2006), does not form part of the current study. An additional item was included to probe participants’ satisfaction with their life as a whole. While not part of the PWI, inclusion of this item facilitated testing for construct validity.

2.2 Participants

A convenience sampling method was used to recruit taxi drivers outside the Beijing railway station. The Beijing railway station represents a major gathering point for Beijing’s taxi drivers with several hundred taxi drivers congregated at any point in time. Taxi drivers were approached in person. Five hundred and twelve Beijing taxi drivers agreed to participate, while 18 taxi drivers declined to be interviewed because they were preparing for a fare.

The characteristics of participants are reported in Table 1. There were 480 males (95.8%) and 21 females (11 participants did not identify their gender), with ages ranging from 20 to 56 (M = 39.65, SD = 6.58). Due to the heavily skewed gender distribution, the psychometric analyses reported in Sect. 3 of this paper pertain only to the males in the sample. On average, the participants work 12.57 h per day and 6.55 days per week. These results indicate that the sample is fairly representative of anecdotal accounts of the Beijing taxi driver population, with a vast majority being married males, earning an income at the lower end of the income spectrum, and working long hours. The descriptive statistics pertaining to gender, marital status and number of children demonstrate limited variability with a majority of the present sample being male, married and having at least one child.

Table 1 Characteristics of participants
Full size table

2.3 Procedure

The PWI was administered in verbal format and was self-completed by participants in 2007. Each participant was assured of anonymity and generally completed the questionnaire containing the PWI and questions about demographic characteristics within 10 min.

2.4 Data Analysis

The data for the male participants were checked for missing values, univariate and multivariate outliers. The Likert scale data were standardized into units of %SM on a 0–100 distribution. Descriptive statistics were used to summarize satisfaction ratings and t-tests and one-way ANOVA were employed to examine the relationship between satisfaction ratings and age, education and income. Cronbach α, item total correlations and item domain correlations were calculated to determine the internal reliability of the PWI. Confirmatory factor analysis was used to confirm the structure of the PWI. Bivariate correlations and multiple regression were conducted to study the inter-relationships between the PWI domains and their contributions to “satisfaction with life as a whole” to establish construct validity.

3 Results

3.1 Internal Reliability of the PWI

3.1.1 Cronbach α and Item-Total Correlations

The Cronbach α coefficient for the PWI is 0.72. The item-total correlations are reported in Table 2. All correlations were significant at the p < 0.01 level and ranged in value between 0.35 and 0.50, with most having a moderate correlation of around 0.45.

Table 2 Item total correlations
Full size table

3.1.2 Domain Inter-Correlations

The domain inter-correlations reported in Table 3 ranged between 0.06 and 0.59. The highest correlations were community connectedness with personal relationships: 0.59; standard of living with life achievement: 0.50 and personal safety with future security: 0.39.

Table 3 Domain Inter-correlations
Full size table

3.2 Validity of the PWI

3.2.1 Factor Analysis

Data on the PWI items were subjected to confirmatory factor analysis under Amos version 7.0 to confirm the unifactorial structure. The independence model confirmed the presence of intercorrelations among the data, indicating the factorability of the scale (χ 228  = 665.96, p < .001). The unifactorial structure of the PWI was confirmed (χ 214  = 55.54, p < .05; CMIN/df = 3.97; GFI = .94; CFI = .93). Factor loadings ranged from .37 to .65 (see Table 4).

Table 4 Confirmatory factor analysis for the PWI items
Full size table

3.2.2 Shared Contributions of Domains to Life as a Whole

The seven domains of the PWI correlate significantly with the general item of “life as a whole”. They ranged from 0.15 to 0.66 (see Table 5). The highest correlations were with standard of living and health. Overall, with the exception of standard of living and health, the correlations are lower than 0.4. To determine the unique contribution of the domains of the PWI to “satisfaction with life as a whole”, the latter was regressed on the former (see Table 5). The model explained in total 50% of the variance. Three domains, namely, standard of living (β = 0.57), personal health (β = 0.21) and personal relationships (β = 0.10) were found to make a significant contribution to life as a whole. The seven domains contribute 8% in unique variance, sharing 42% of variance between them.

Table 5 Regression of ‘satisfaction with life as a whole’ on personal domains
Full size table

3.3 Satisfaction Ratings of the PWI

The means and standard deviations of the domains of the PWI for the male participants are given in Table 6. The mean domain scores ranged from 44.4 (SD = 27.4) to 77.4 (SD = 21.1) and the PWI score was 61.2 (SD = 15.2). Satisfaction with life achievement, personal relationships and community connectedness lie above the PWI score, while standard of living, health personal safety and future security lie below the PWI score. The highest mean score was for satisfaction with personal relationships (77.4, SD = 21.2). The lowest mean score was for future security (44.4, SD = 27.4).

Table 6 Satisfaction ratings of the PWI
Full size table

3.3.1 Work Hours and Wellbeing

The correlation between mean hours worked per day and wellbeing was −.12 (p < .05). The correlation between means days worked per week and wellbeing was −.10 (p < .05).

3.3.2 Age and Wellbeing

Table 7 presents participants’ satisfaction ratings broken down according to three age categories: 18–35, 36–45 and 46–60. The highest mean score was reported by the age group 46–60: 62.0 (SD = 16.0), while the lowest mean scores was reported by the youngest age group, 18–35: 60.4 (SD = 15.7). There was, however, little difference in the mean scores between age groups and these differences were not significant.

Table 7 Personal wellbeing and age
Full size table

3.3.3 Education and Wellbeing

Table 8 presents participants’ satisfaction ratings broken down according to education. Those with a technical or bachelor degree had the highest PWI score: 61.7 (SD = 14.6); but there was little difference in satisfaction as a function of education and differences in PWI scores were not statistically significant with respect to education.

Table 8 Personal wellbeing and education
Full size table

3.3.4 Income and Wellbeing

Table 9 presents participants’ satisfaction ratings broken down according to income. Higher income earners reported higher personal wellbeing. A one-way ANOVA found differences across income categories were significant. Tukey’s HSD tests confirmed that generally higher income categories reported higher personal wellbeing than lower income categories across the board, except for the two highest income categories, where no significant difference in personal wellbeing was found.

Table 9 Personal wellbeing and income
Full size table

4 Discussion

The present study has provided data about subjective well-being among a sample of male Beijing taxi drivers. The study adds to the literature on subjective well-being in China for a demographic profile that has until now not been studied. The aims were to (a) investigate the psychometric properties of the PWI; (b) ascertain whether Beijing taxi drivers are satisfied with their lives; and (c) examine whether the responses to the PWI from taxi-driver participants falls within the narrow range predicted by the “Theory of Subjective Wellbeing Homeostasis”. The findings are discussed below in relation to these aims.

4.1 Psychometric Properties of the PWI

The first objective of the present study was to examine the psychometric properties of the PWI in a new demographic group. This purpose is important given that, as Chen and Davey (2009) noted in the quotation reproduced in the introduction, the PWI remains a work in progress, particularly with Chinese samples for which the empirical evidence is limited. The PWI was found to exhibit good reliability, validity and sensitivity as a measure of subjective well-being. This result concurs with the findings from existing studies which have applied the PWI to Chinese samples (Chen and Davey 2009; Lau et al. 2005, 2008) and samples from Western countries (International Wellbeing Group 2006), which have concluded the PWI has good psychometric properties. The Cronbach α coefficient for the PWI demonstrates good reliability and is generally comparable to the findings of earlier studies. For example, Lau et al. (2005) found the Cronbach α coefficient in a sample of Australian participants to be 0.73 and Davey et al. (2009) found the Cronbach α coefficient in a study of rural China to be 0.75. The item-total correlations are similar in magnitude to, albeit slightly lower than, the findings reported in previous studies. For example, Lau et al. (2005) found that the majority of items in Australian and Hong Kong samples had a correlation of around 0.50.

The seven domains of the PWI were found to correlate significantly with the general item of “life as a whole”. The highest correlations were with standard of living and health, which is similar to findings from previous studies (see e.g. Lau et al. 2005 for Australia, Tiliouine et al. 2006 for Algeria). The results for the other domains were similar to what Nielsen et al. (2009) found for Chinese off-farm migrants and Smyth et al. (2009) found for Chinese urban residents across six cities, but, in general, were lower than those found in most previous studies (see e.g. Cummins et al. 2003, 2004; Lau et al. 2005; Tiliouine et al. 2006).

The results for the unique contribution of the domains of the PWI to “satisfaction with life as a whole” was similar to previous studies for Australia (43%) (Lau et al. 2005), Zhuhai in Guangdong province (47%) (Chen and Davey 2009); six Chinese cities (46%) (Smyth et al. 2009) and Chinese off-farm migrants (44%) (Nielsen et al. 2009). Standard of living, personal health and personal relationships were found to make a significant contribution to life as a whole. Previous studies have also found standard of living to make the largest unique contribution to predicting life as a whole (Nielsen et al. 2009; Lau et al. 2005; Renn et al. 2009; Smyth et al. 2009; Tiliouine et al. 2006). The unique and shared variance explained is similar to the findings from previous studies for China. Chen and Davey (2009) found that total explained unique variability was 12% and shared variability was 36% in Zhuhai in Guangdong province. Smyth et al. (2009) found that unique variability was 14% and shared variability was 32% in six Chinese cities. Nielsen et al. (2009) found that total explained unique variability was 16% and shared variability was 29% for a sample of Chinese off-farm migrants in Fujian province.

4.2 Satisfaction with Life

The PWI scores show that male taxi drivers in Beijing, on average, were satisfied with their lives. The PWI score and, apart from future security, the domain scores were situated above the scale midpoint, which is indicative of a moderate, positive level of well-being (Chen and Davey 2009; Nielsen et al. 2009). These results are consistent with existing findings for Mainland China (Chen and Davey 2008a, 2009; Davey et al. 2009; Huang and Xing 2005; Nielsen et al. 2009; Smyth et al. 2009), other Chinese societies (Chen and Davey 2008b; Lau et al. 2005, 2008; Macau Inter-University Institute 2007) and studies for Western countries employing the PWI and other instruments to measure well-being, that most people are content with their lives (International Wellbeing Group 2006).

The lowest satisfaction scores were reported for future security: 44.4 (SD = 27.4) and personal safety: 53.9 (SD = 26.1). The results for future security reflect several factors. First, for the vast majority of Beijing taxi drivers, who are company drivers, the employment contract they have with the taxi companies gives them no job security. These drivers are reluctant to speak out in favour of better wages and shorter working hours for fear of losing their jobs. Second, at the time the survey was conducted in 2007, there had been several increases in the price of taxi fares reflecting the rapid increase in the price of oil and these price increases were expected to increase. The increase in taxi fares was adversely affecting Beijing commuters, who were looking for alternative means of transportation. Third, the Beijing taxi industry is changing rapidly in response to environmental pressures, which brings uncertainty for the drivers. The Beijing taxi industry was under pressure to move towards adopting cleaner vehicle technologies in time for the Beijing Olympics (Chongfang et al. 2004; Zhao 2006).

Safety among Beijing taxi drivers is an important area for concern, considering recent rates in traffic accidents. Officially, traffic accidents in Beijing accounted for 1.53% of total traffic accidents in China in 2006. This equates to a total of 5,808 traffic accidents in Beijing in 2006 (State Statistical Bureau 2007). Official statistics on traffic accidents, however, are most likely limited to those that have either been reported or have warranted police intervention, and so may in fact under-represent cases of traffic accidents in Beijing.

4.3 Applicability of the ‘Theory of Subjective Wellbeing Homeostasis’

The third objective was to ascertain whether the responses to the PWI from participants fell within the narrow range predicted by the “Theory of Subjective Wellbeing Homeostasis”. That the PWI displayed good psychometric properties, and that the PWI fell within the 60–70%SM range predicted for non-Western countries, is consistent with the “Theory of Subjective Wellbeing Homeostasis”. The positioning of the Beijing taxi driver data within this range supports the proposition that a psychological homeostatic mechanism is in operation. While, based on objective indicators, the quality of life of Beijing taxi drivers is lower than the urban adult population as a whole in Mainland China and certainly the adult population in Hong Kong and Macau, their subjective well-being is in the same normative range. The results reported here, together with the findings for Chinese off-farm migrants reported in Nielsen et al. (2009) suggest that the homeostatic mechanism is fairly resilient, even when the individual leads a hard life based on objective indicators.

Nielsen et al. (2009) conjectured that the circular nature of migration was a release valve which provided a buffer on the homeostatic mechanism for Chinese off-farm migrants. Beijing taxi drivers do not have this release valve; however, it is likely that Beijing taxi drivers have other external buffers exist which can assist the homeostatic mechanism to prevent subjective wellbeing falling below the threshold value. Two external buffers, which have been identified as being particularly important are personal relationships and income.

Male Beijing taxi drivers were most satisfied with personal relationships and community connectedness, which are fundamental to the construal of the self in Chinese culture (see Markus and Kitayama 1991). Most previous studies have also found that personal relationships are the part of their lives with which they are most satisfied. Cummins et al. (2007) noted that personal relationships act as a powerful external buffer on the homeostatic mechanism. This observation is particularly true for demographic groups such as the Beijing taxi drivers in the present study and off-farm migrants in the study by Nielsen et al. (2009) for which the PWI is at the lower end of the normative range predicted by the “Theory of Subjective Wellbeing Homeostasis”. If these demographic groups, which otherwise lead hard lives, did not have a strong network of personal relationships to fall back on it is likely that subjective homeostatic well-being would be defeated. In this sample a 10.2% drop in mean satisfaction with personal relationships, all things being equal, would reduce the PWI below the lower bound of the normative range predicted by the theory for non-Western societies.

Connectedness to the community reinforces personal relationships in assisting the homeostatic mechanism. That Beijing taxi drivers reported a high satisfaction score for community connectedness is not surprising. As Markus and Kitayama (1991) pointed out, in several non-western cultures, including Chinese culture, the self is viewed as interdependent with the surrounding context and as such, construals of the self are made in relation to others, rather than with a self-focus. The manifestations of such interdependent construals of the self are, for example, that consideration and expression of motives and emotions are governed by a consideration of the reactions of others. The nature of the occupation, together with the number of hours spent in their taxis, means that Beijing taxi drivers are constantly coming into contact with a broad cross-section of the community who are getting into their cabs. Most people travelling in taxis engage in casual conversation about the events of the day. Moreover, many taxi drivers have the radio turned on while in their taxis, which carries news bulletins with information on local events, which are then, in turn, discussed with passengers. Thus, taxi drivers are uniquely placed to feel part of the local community.

That generally higher income categories reported higher personal wellbeing than lower income categories across the board, except for the two highest income categories is consistent with previous findings for urban China (Smyth et al. 2009). This finding is consistent with the view that income is a flexible resource which can be used to assist homeostasis (Cummins 2000). Income allows people to minimize the unwanted challenges they encounter in their daily lives. Higher income individuals can pay others to do tasks that they do not want to perform (Cummins et al. 2009). For this reason, higher income has been shown to be associated with higher subjective well-being in a cross-sectional sample, as in the present study, at a single point in time. This is true for western samples (see Clark et al. 2008; Dolan et al. 2008 for reviews) and samples from rural and urban China (see e.g. Appleton and Song 2008; Knight et al. 2009; Smyth et al. 2009). However, there is much evidence that well-being does not increase with income over time (Easterlin 1974). Irrespective of how rich one becomes, their subjective well-being cannot be sustained higher than the set-point range over time. The lowest income earners, who lack the financial resources, must fend for themselves to a greater extent and, as a consequence, their subjective well-being is more threatened by external threats to the homeostatic mechanism. It is interesting to note that the lowest income group has PWI scores below the normative range of 60–70 predicted by the “Theory of Subjective Wellbeing Homeostasis”.

In concluding it is important to note that the present study contains some limitations. While it has been argued that focusing on Beijing taxi drivers, who work under harsh conditions, is a worthwhile exercise because this group represents a strong test for “Theory of Subjective Wellbeing Homeostasis”, the results need not be generalizable to other groups. Moreover, correlations with potential predictors of well-being, such as demographic variables, may be reduced in a highly uniform sample, such as this. Further research is needed to ascertain whether the results are generalizable to other locales and occupations. Second, the survey did not include satisfaction with religion/spirituality, which has only been one of the domains of the PWI since November, 2006 (International Wellbeing Group 2006). Third, generally studies have shown that socio-demographic attributes explain little variance in subjective well-being. For this reason, it might be more fruitful in future research to investigate the interplay between significant aspects of personality, such as personal goals, traits, and self-construal, and well-being, particularly domain-specific well-being.