Reliability and validity of the Japanese version of the Survey of Perceived Organizational Support

Abstract

Objectives

This study aimed to examine the reliability and validity of the Japanese version of the Survey of Perceived Organizational Support (SPOS-J) for the Japanese workforce.

Methods

The translation of the SPOS into Japanese followed the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) Task Force guidelines. An online baseline survey with 6220 Japanese workers assessed the SPOS-J and related scales, followed by a follow-up survey with 452 respondents 2 weeks later. We developed the SPOS-J and evaluated its factorial, convergent, and known-group validities, as well as its internal consistency, test–retest reliability, and interpretability.

Results

Exploratory and confirmatory factor analysis (EFA and CFA) indicated that the SPOS-J, diverging from the original version’s 1-factor structure, adopted a 2-factor structure comprising 2 subscales: the SPOS-J (Sufficiency of Positive Aspects [SPA]) and the SPOS-J (Minimality of Negative Aspects [MNA]). The final version of the SPOS-J, consisting of 8 items selected by the item response theory analysis, demonstrated acceptable model fit in the CFA. Cronbach a values for SPOS-J (SPA) and SPOS-J (MNA) were .92 and .84; intraclass coefficients were 0.72 and 0.55, respectively. The correlations between SPOS-J (SPA) and related variables, as well as the score distributions by demographic characteristics, supported the convergent and known-group validity of SPOS-J (SPA). In contrast, SPOS-J (MNA) showed lower correlation coefficients with all related variables and supported known-group validity.

Conclusions

The SPOS-J (SPA), a subscale of the 8-item SPOS-J, is a reliable and valid measure for assessing perceived organizational support among Japanese workers.

Key points

• What is already known on this topic: Because there is currently no Japanese version of the SPOS that has been validated for reliability and validity, studies in Japan are mainly conducted using original questionnaires that reflect the POS concept.

• What this study adds: Our study provides a reliable and validated POS measurement scale in Japan based on the original version of SPOS.

• How this study might affect research, practice, or policy: The implementation of SPOS-J as a measure for evaluating activities aimed at enhancing employees’ health and well-being allows for introducing a new evaluative metric for occupational health practices and health and productivity management within Japanese companies.

Introduction

Perceived organizational support (POS) is a concept proposed by Eisenberger et al¹ in 1986, and is defined as employees’ “global beliefs concerning the extent to which the organization values their contributions and cares about their well-being.” A major characteristic of POS is that it focuses not on the company’s (management and human resource management department’s) point of view, but on the employees’ point of view, that is, how the employees who work there feel about the company. POS is based on social exchange theory, the theory that human beings form and develop relationships through the exchange of various tangible and intangible rewards.¹ Therefore, it is based on the idea that the more support an employee feels from the organization, the more favorable the employee feels toward the organization, and the more the employee will make further efforts to contribute to the achievement of the organization’s goals. Therefore, POS is seen as a concept with important benefits for both employees and organizations; research on POS began to increase in the mid-1990s and has progressed rapidly in recent years.

Studies have shown a relationship between POS and a variety of health-related indicators. Antecedents of POS include procedural justice² (ie, one of the subconcepts of organizational justice), supervisor support,^3,4 organizational rewards,⁵ and job conditions,^6–8 whereas outcome factors include emotional commitment to the organization,³ work engagement,⁹ job satisfaction,¹⁰ and performance.^11,12 Many of these antecedent and outcome factors are positively correlated with POS. On the other hand, antecedent factors such as job conditions (role ambiguity,⁷ role conflict,⁸ size of organization⁷) and organizational politics,¹³ and outcome factors such as withdrawal behavior (turnover intentions,¹⁴ turnover⁴) and strains are negatively correlated with POS.

A representative measure of POS is the Survey of Perceived Organizational Support (SPOS).¹⁵ As Eisenberger and Stinglhamber¹⁶ note, “Over 95% of POS studies on perceived organizational support use the Survey of Perceived Organizational Support,” indicating that the SPOS is the exclusive scale for capturing POS. The SPOS is a 36-question measure developed by Eisenberger et al in 1986.¹ In much of the past literature, the original version of the SPOS was 1-dimensional and had high internal reliability, leading to the use of shortened versions such as the 16-item version¹ and the 8-item version.⁶ Studies using the SPOS have been conducted mainly in English-speaking countries and European countries, although besides the English version, Malay¹⁷ and Turkish¹⁸ versions of the SPOS have been developed. However, there is currently no Japanese version of the SPOS that has been tested for reliability and validity; therefore, studies on POS with Japanese employees have been conducted using original questionnaire items that reflect the POS concept.¹⁹

Currently, in many developed countries, including Japan, the working-age population is declining due to the low birthrate and aging population.²⁰ Therefore, companies must make strategic efforts to improve POS by actively considering the well-being of their employees. This aims to enhance corporate performance and reduce employee turnover by increasing job satisfaction and work engagement. Furthermore, as companies need to build good relationships with diverse employees due to the recent diversification of employment patterns and needs, concepts and indicators that reflect employees’ evaluations and opinions of the company, such as POS, are useful in verifying the effectiveness of personnel policies aimed at embracing diversity.

The purpose of this study was to examine the reliability and validity of the Japanese version of the SPOS (SPOS-J) for the general workforce in Japan. Specifically, we assessed its factorial, convergent, and known-group validities as well as internal consistency, test–retest reliability, and interpretability.

Methods

Participants

A baseline online survey was conducted from January 24 to February 3, 2022, among registrants of Cross Marketing Inc. (Tokyo, Japan), a private online survey firm in Japan. As of January 2022, there were approximately 2.95 million registered participants, from which 102 032 were randomly selected to receive participation invitations via email. Of these, 15 676 responded to the preliminary survey and participated in the main survey (baseline survey) if they met all of the screening criteria: (1) age 20-64, (2) currently working, and (3) employed (full-time, contract, temporary, part-time, or casual employee). For this baseline survey, the sample was collected based on data from the 2020 Labor Force Survey (Ministry of Internal Affairs and Communications), stratified by gender (men: 54.4%, women: 45.6%), age (20-29: 18.6%, 30-39: 21.1%, 40-49: 27.7%, 50-59: 24.0%, 60-64: 8.7%), and region of residence (Hokkaido: 3.8%, Tohoku: 6.5%, South Kanto: 31.6%, North Kanto/Koshin: 7.4%, Hokuriku: 4.1%, Tokai: 12.2%, Kinki: 15.7%, Chugoku: 5.5%, Shikoku: 2.6%, Kyushu/Okinawa: 10.5%). Responses to this survey were on a first-come, first-served basis, and due to budget constraints, the target number of respondents was set at 6000. Once the final 6220 respondents were reached, the recruitment process was closed because the sample was collected at a stratified percentage.

In addition, to examine the test–retest reliability of the SPOS-J, a follow-up study using the SPOS-J was conducted 2 weeks after the baseline study. The follow-up survey was administered to those respondents of the baseline survey who answered “no” to both (1) “changed workplaces within the past two weeks” and (2) “had a significant change in workplaces within the past two weeks” at the time of their follow-up survey responses. As in the baseline survey, responses were on a first-come, first-served basis, and the survey was closed when the number of respondents reached 452. Because the online survey required all questions to be answered, no participants had incomplete responses.

Measures

SPOS-J

The Japanese version of the SPOS was prepared in accordance with the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) Task Force guidelines.²¹ First, permission for the Japanese translation was obtained from the original author (Robert Eisenberger) (Step 1: Preparation), and the original version of SPOS¹ was independently translated into Japanese by 4 native Japanese-speaking researchers (K.O., T.N., H.E., and K.Mori.) (Step 2: Forward Translation), and a draft Japanese version was prepared after discussion among the researchers (Step 3: Reconciliation). Next, native experts who had not read the original text were asked to translate the draft Japanese version into English (Step 4: Back Translation). The original author reviewed the back-translated English version, and the draft Japanese version was revised. Because there is no corresponding Japanese word for “well-being” and the specific image of “well-being” varies from person to person, we consulted with the original author and added the definition from the Longman Dictionary of Contemporary English,²² adding that in the context of work it means “a feeling of being comfortable, healthy, and happy through work” as a supplementary explanation (Step 5: Back Translation Review; Step 6: Harmonization). Next, as a cognitive debriefing, we asked 5 workers of various occupations and ages collected through snowball sampling for their opinions by asking them, “Please advise us if you have any difficulty understanding or answering any questions in Japanese” (Step 7: Cognitive Debriefing). As a result, minor comments were received, including suggestions for the use of more easily understood verbs and adjectives. The 4 researchers considered the pros and cons of reflecting these comments in the translation, and adjusted the translation so that it would not affect the meaning when translated into English (Step 8: Review of Cognitive Debriefing Results and Finalization; Step 9: Proofreading; Step 10: Final Report). The Japanese version of SPOS (SPOS-J) was thus completed (Table SS1).

The SPOS-J consists of 36 items, and as in the original version of the SPOS, respondents were asked to respond to each item on a 7-point scale (0 = strongly disagree to 6 = strongly agree). Of the 36 items, 18 items asked for positive content, whereas the remaining 18 items asked for negative content. The scoring of the latter items asking about negative content was reversed, with higher scores for all items indicating a higher POS. Finally, the average score for the 36 items was calculated.

Relevant variables

In this study, the following indicators were selected based on existing studies that have shown a strong correlation with POS¹⁵ and the availability of valid Japanese language scales. Role conflict from job conditions and turnover intention from withdrawal behavior were selected as indicators of negative correlation with POS.

Possible antecedents

• Supervisor support: The corresponding subscale of the New Brief Job Stress Questionnaire (New BJSQ) was used. The New BJSQ has been validated with Japanese workers and found to have acceptable levels of internal consistency, reliability, and construct validity on almost all subscales.²³ Three items from the New BJSQ regarding supervisor support (eg, How freely can you talk with your supervisor?) were rated using a 4-point scale (1 = extremely to 4 = not at all). The scores were reversed in the analysis to make them more intuitive and easier to understand, and the average score for the 3 items was calculated. The Cronbach a coefficient for this scale was .88.

• Organizational justice (procedural justice and interactional justice): Using the Japanese version of the Organizational Justice Scale,²⁴ which is the Japanese version of the modified version of the Organizational Justice Scale²⁵ developed by Moorman,²⁶ we rated each of the 2 subscales, procedural justice (7 items) and interactional justice (6 items), using a 5-point scale (1 = strongly disagree to 5 = strongly agree) and calculated the mean score. Cronbach a coefficients for these subscales were .94 and .96 for procedural and interactional justice, respectively.

• Role conflict: One item on the New BJSQ (“I receive incompatible requests from two or more people”) was rated using the 4-point scale (1 = definitely to 4 = not at all), and the scores were reversed in the analysis for easier understanding.

Possible outcomes

• Job satisfaction: One item (“I am satisfied with my job”) from the New BJSQ was used to evaluate the results. This indicator was rated using a 4-point scale (1 = satisfied to 4 = dissatisfied), and the scores were reversed in the analysis for ease of understanding.

• Work engagement: Work engagement was assessed using the 3-item Japanese version of the Utrecht Work Engagement Scale (UWES).^27–29 The UWES includes 3 subscales (vigor, dedication, and absorption), with each subscale being evaluated on a 7-point scale (0 = never to 6 = always). The average score of the 3 items was calculated. The Cronbach a coefficient for this scale was .93.

• Organizational (affective) commitment: Three items from the Japan Institute of Labour (currently The Japan Institute for Labour Policy and Training) Human Resource Management checklist³⁰ regarding affective commitment (eg, “I feel as if this company’s problems are my own problems”) were evaluated using a 5-point scale (1 = not at all true to 5 = very true), and the average score was calculated. The Cronbach a coefficient for this scale was .73.

• Turnover intention: A 1-question shortened version of the Japanese version of the Turnover Intention Scale³¹ (“How often have you seriously considered quitting your current job?”) was used and rated using a 6-point scale (1 = not at all to 6 = very often).

Demographic characteristics

Participant demographic items were collected at the time of the baseline survey. In addition to gender (men, women) and age, respondents were asked to provide their final education (junior high school/high school, vocational school/junior college, university/graduate school), household income (<3.00 million Japanese yen [JPY], 3.00-4.99 million JPY, 5.00-9.99 million JPY, and 10.00 million JPY or greater), organizational size (1-99 employees, 100-299 employees, 300-499 employees, 500-999 employees, 1000 employees or larger, and public officers), and job type (administrative, professional/technical, clerical, sales, service, security, agriculture/forestry/fishing, production process jobs, transportation/machine operation, construction/mining, transportation/cleaning/packaging, and others) based on the Japan Standard Occupational Classification (statistical standard established in December 2009).

Statistical analysis

A confirmatory factor analysis (CFA) was performed using a 1-factor model for SPOS-J based on the original version of the SPOS¹ by Eisenberger et al. Model fit was evaluated using a combination of goodness-of-fit indices, including χ² goodness-of-fit statistic, root mean square error of approximation (RMSEA), Akaike information criterion (AIC), comparative fit index (CFI), and standardized root mean squared residual (SRMR). Acceptable model fit criteria were defined as RMSEA <0.06, SRMR <0.08, and CFI >0.95.³² As will be detailed later, because the CFA showed a poor fit for the 1-factor model, an exploratory factor analysis (EFA) was subsequently conducted. An EFA (maximum likelihood method, determining the number of factors with the smallest eigenvalue as 1,³³ promax rotation) was conducted on all 36 items of the SPOS-J. Among each item, items with factor loadings of less than 0.4 and items with factor loadings greater than 0.4 across multiple factors were deleted.³⁴ A CFA was performed on the factor structure supported by the EFA, and the model’s goodness of fit was compared with the 1-factor structure of the original version. Due to the unacceptable model fit of the factor structure supported by the EFA, we attempted to reduce the SPOS-J items by conducting an item response theory (IRT) analysis with the generalized partial credit model (GPCM).³⁵ The discrimination (a) and difficulty (b/threshold) of each item were estimated, and we selected items for each factor based on the excellence of these parameters. A CFA using the maximum likelihood estimation method was conducted to confirm the factor structure of the items finally selected as SPOS-J.

To assess test–retest reliability, the Kolmogorov-Smirnov test was first used to check whether the total scores for each of the SPOS-J subscales follow a normal distribution. The results indicated that the scores did not follow a normal distribution. However, upon visually inspecting the data distribution using histograms and normal Q-Q plots, it appeared to closely approximate a normal distribution. In large sample sizes, the Kolmogorov-Smirnov test is known to detect statistically significant deviations even with slight deviations,^36,37 emphasizing the recommendation to interpret results alongside visual verification. Therefore, based on the visual confirmation that the data closely approximate a normal distribution, we deemed it appropriate to treat the total scores of SPOS-J subscales as parametric variables rather than nonparametric variables. As a result, in this study, we calculated Pearson correlations, intraclass correlation coefficient (ICC [2,1]), and standard error of measurement (SEM) to assess test–retest reliability. Following Cohen,³⁸ we describe effects as small (0.10), medium (0.30), and large (0.50). Data from 452 individuals who responded to the follow-up survey were used to calculate the ICC and SEM.

To assess convergent validity, Pearson correlation coefficients were calculated for the SPOS-J subscales with the antecedents (supervisor support, organizational justice, and role conflict) and outcome factors (job satisfaction, work engagement, organizational commitment, and turnover intention). To assess internal consistency, Cronbach a coefficients were calculated for the SPOS-J subscales.

To assess known-group validity, a 1-way analysis of variance (ANOVA) and Tukey multiple comparison test were conducted to investigate whether there were significant differences in mean scores of the SPOS-J subscales across demographic characteristics (age and organizational size). In international studies, a weak positive correlation has been shown between POS and age.¹⁵ Regarding the relationship between POS and organizational size, it has been reported that in companies with up to 500 employees, POS negatively correlates with organizational size.⁷ Additionally, Inoue et al³⁹ found that in their study of Japanese workers, POS was lowest in medium-sized companies (50-999 employees) and higher in both small (2-49 employees) and large (1000 or more employees) companies compared with medium-sized ones. Therefore, we hypothesized that the mean scores of each SPOS-J subscale would increase with age and decrease with organizational size up to a certain point, then increase as organizational size becomes larger.

Finally, to assess interpretability, we confirmed the minimal clinically important difference (MCID). MCID is defined as “the threshold above which clinicians, patients, and researchers perceive an outcome difference.”⁴⁰ Determining the MCID allowed us to understand the clinical significance of changes measured by the scale. There are 2 main approaches to determining the MCID: anchor-based and distribution-based.⁴¹ The anchor-based approach compares changes in outcome measures with changes in an external criterion (anchor), but because the anchor we are measuring does not have a clear cutoff point, we used the distribution-based approach. We adopted the SD method, using 0.5 SD of the baseline score as the MCID. Although various cutoff values are used in the SD method, there is a consensus that a change of 0.5 SD is clinically significant.⁴²

The significance level was set at .05 (2-tailed). All analyses were performed using IBM SPSS Statistics Version 27.0, IBM SPSS Amos Version 27.0, and Stata version 16.

Results

Characteristics of participants

Table 1 shows the number of participants in the baseline and follow-up surveys and the characteristics of the participants. In the baseline survey, participants had the same gender and age distribution as in the data from the 2020 Labor Force Survey (Ministry of Internal Affairs and Communications), participants’ annual household income of 5.00-9.99 million JPY was the most prevalent (43.2%), and 50.5% of participants had a college degree or higher as their final education level. The organizational size was predominantly small (1-99 employees) or large (>1000 employees), and the most common occupations were clerical, professional/technical, and service, in that order. There were no significant differences in gender, age, final education, organizational size, or occupation between baseline and follow-up.

Table 1.

Demographics of participants at baseline and follow-up survey.

	Baseline		Follow-up a
Demographic characteristics	n (%)		n (%)
Total	6220		452
Gender
Men	3382	(54.3)	248	(54.9)
Women	2838	(45.6)	204	(45.1)
Age, y
20-29	1131	(18.2)	59	(13.1)
30-39	1281	(20.6)	66	(14.6)
40-49	1684	(27.1)	142	(31.4)
50-59	1455	(23.4)	108	(23.9)
60-64	669	(10.8)	77	(17.0)
Annual household income (million yen)
<3.00	1073	(17.3)	74	(16.4)
3.00-4.99	1658	(26.7)	128	(28.3)
5.00-9.99	2689	(43.2)	185	(40.9)
³10.00	800	(12.9)	65	(14.4)
Final education
Junior high or high school	1748	(28.1)	128	(28.3)
Vocational school or college	1331	(21.4)	91	(20.1)
University or graduate school	3141	(50.5)	233	(51.6)
Organizational size (employees)
1-99	2382	(38.3)	179	(39.6)
100-299	892	(14.3)	71	(15.7)
300-499	371	(6.0)	18	(4.0)
500-999	492	(7.9)	33	(7.3)
³1000	1810	(29.1)	129	(28.5)
Public offices	273	(4.4)	22	(4.9)
Job type
Administrative	402	(6.5)	35	(7.7)
Professional and technical	1095	(17.6)	67	(14.8)
Clerical	1702	(27.4)	123	(27.2)
Sales	616	(9.9)	46	(10.2)
Service	842	(13.5)	54	(12.0)
Security	91	(1.5)	7	(1.6)
Agriculture, forestry, and fishing	16	(0.3)	3	(0.7)
Production process jobs	459	(7.4)	42	(9.3)
Transportation and machine operation	115	(1.9)	8	(1.8)
Construction and mining	91	(1.5)	4	(0.9)
Transportation, cleaning, packaging	217	(3.5)	19	(4.2)
Others	574	(9.2)	44	(9.7)

^aFollow-up sample was used only to examine the test–retest reliability

Factorial validity

Table 2 shows factor structures, discrimination (a), and difficulty (b) of the SPOS-J items. Based on the original version of the SPOS, a CFA was performed on the SPOS-J with a 1-factor model, and the model fit was not acceptable (χ² = 85 197; P < .001; RMSEA = 0.151 [90% CI, 0.150 to 0.152), SRMR = 0.245, CFI = 0.511, and AIC = 85 341), so an EFA was performed. The EFA yielded initial eigenvalues of 11.498, 9.524, 1.807, 1.053, 0.764, and 0.655 for factors 1 through 6, and the number of factors was determined to be 4 according to the Guttman standard. Items with factor loadings of 0.4 or higher were retained, leading to the exclusion of any items from factors 3 and 4. Furthermore, 3 items (Item nos.14, 22, and 26) with loadings above 0.4 across multiple factors were excluded. Consequently, 33 items were retained, and 2 factors, Factor 1 and Factor 2, were considered appropriate as dimensions for measurement. The first factor involved items assessing the sufficiency of organizational support (eg, “My organization values my opinion”), whereas the second factor concerned items evaluating the minimality of nonsupport (eg, “My organization shows little interest in my interests”). Hence, the first factor was labeled SPOS-J (Sufficiency of Positive Aspects [SPA]), and the second SPOS-J (Minimality of Negative Aspects [MNA]). Subsequent CFA on the 33 remaining items aimed to assess the fit of the 2-factor model, which remained unacceptable (χ² = 18 492; P < .001, RMSEA = 0.077 [90% CI, 0.076 to 0.077), SRMR = 0.067, CFI = 0.885, AIC = 18 626). This prompted further IRT analysis for each dimension, identifying discrimination and difficulty parameters of the items. After excluding items with reversed transition points for difficulty, 4 items remained as SPOS-J (MNA). In the development of affective scales, it is customary to include an equal number of positive and negative questions to mitigate response bias.^17,43 Thus, the top 4 items were chosen as SPOS-J (SPA) based on their increasing discriminability, ensuring there was no reversal in the difficulty level among these items. The SE for estimating latent factors (θ) was 0.012 for both of the 2 dimensions. A total of 8 items (4 items for each of the 2 factors) were selected as items for the final version of the SPOS-J. An additional CFA on these 8 items indicated an acceptable fit for the 2-factor model (χ² = 403; P < 0.001, RMSEA = 0.057 [90% CI, 0.052 to 0.062], SRMR = 0.015, CFI = 0.988, AIC = 436).

Table 2.

Exploratory factor analysis (standardized factor loading of items in the 2-factor structure, interfactor correlations, and eigenvalues among factors) and item response theory analysis (discrimination and difficulty of items and SE for theta in each factor) of the SPOS-J (n = 6220).

		Exploratory factor analysis (EFA)		Item response theory (IRT) analysis
		Standardized factor loading			Difficulty (b, threshold)						SE for theta (θ)	Adopted item
Item no.	Item in English	Factor1	Factor2	Discrimination (a)	b1 (1 vs 0)	b2 (2 vs 1)	b3 (3 vs 2)	b4 (4 vs 3)	b5 (5 vs 4)	b6 (6 vs 5)
25	The organization cares about my opinions.	0.876	0.029	2.597	−1.528	−0.971	−0.526	0.618	1.305	2.127	0.012	✓
27	The organization takes pride in my accomplishments at work.	0.863	0.014	1.958	−1.408	−0.915	−0.654	1.109	1.516	2.329
36	My supervisors are proud that I am a part of this organization.	0.862	0.060	1.863	−1.225	−0.827	−0.686	1.180	1.476	2.289
21	The organization cares about my general satisfaction at work.	0.859	0.061	2.339	−1.283	−0.802	−0.419	0.966	1.518	2.291		✓
10	The organization is willing to extend itself in order to help me perform my job to the best of my ability.	0.856	0.034	2.647	−1.373	−0.847	−0.369	0.862	1.425	2.155		✓
9	The organization really cares about my well-being.	0.842	0.041	2.430	−1.267	−0.825	−0.360	0.975	1.446	2.280		✓
20	The organization is willing to help me when I need a special favor.	0.832	-0.049	1.909	−1.534	−0.972	−0.718	0.866	1.463	2.179
33	The organization wishes to give me the best possible job for which I am qualified.	0.810	-0.043	1.604	−1.444	−0.988	−0.810	1.155	1.564	2.309
8	Help is available from the organization when I have a problem.	0.805	-0.026	1.829	−1.502	−0.918	−0.550	0.798	1.303	2.287
4	The organization strongly considers my goals and values.	0.797	0.026	1.763	−1.354	−0.835	−0.473	0.996	1.481	2.263
1	The organization values my contribution to its well-being.	0.777	0.053	1.557	−1.301	−0.762	−0.394	0.786	1.294	2.567
35	The organization tries to make my job as interesting as possible.	0.761	0.081	1.200	−0.938	−0.765	−0.587	1.885	1.922	2.559
24	If I decided to quit, the organization would try to persuade me to stay.	0.745	0.004	0.892	−1.122	−0.962	−1.176	1.649	1.433	2.206
29	The organization would understand if I were unable to finish a task on time.	0.739	-0.129	1.052	−1.504	−1.215	−1.060	1.259	1.570	2.406
13	The organization would forgive an honest mistake on my part.	0.737	-0.170	1.102	−1.698	−1.218	−1.117	0.765	1.243	2.291
18	The organization would grant a reasonable request for a change in my working conditions.	0.723	-0.037	1.123	−1.195	−0.886	−0.847	1.747	1.680	2.685
30	If the organization earned a great profit, it would consider increasing my salary.	0.709	0.086	0.910	−0.644	−0.650	−0.853	1.860	2.035	2.836
5	The organization would understand a long absence due to my illness.	0.676	-0.122	0.879	−1.437	−1.187	−1.086	1.003	1.029	1.952

Table 2.

Continued

		Exploratory factor analysis (EFA)		Item response theory (IRT) analysis
		Standardized factor loading			Difficulty (b, threshold)						SE for theta (θ)	Adopted item
Item no.	Item in English	Factor1	Factor2	Discrimination (a)	b1 (1 vs 0)	b2 (2 vs 1)	b3 (3 vs 2)	b4 (4 vs 3)	b5 (5 vs 4)	b6 (6 vs 5)
17	Even if I did the best job possible, the organization fails to notice.	0.103	0.838	1.987	−1.337	−1.031	−0.908	0.481	0.933	1.830	0.012	✓
23	The organization feels there is little to be gained by employing me for the rest of my carrier.	0.066	0.799	1.713	−1.408	−1.247	−1.165	0.566	0.943	1.678		✓
32	The organization is unconcerned about paying me what I deserve.	0.026	0.780	1.430	−1.140	−1.144	−1.261	0.800	1.079	1.923
3	The organization fails to appreciate any extra effort from me.	0.051	0.761	1.412	−1.353	−1.228	−1.033	0.478	0.914	1.823		✓
7	The organization disregards my best interests when it makes decisions that affect me.	0.055	0.755	1.580	−1.703	−1.295	−1.256	0.468	0.944	1.841		✓
16	The organization provides me little opportunity to move up the ranks.	−0.001	0.741	0.950	−0.389	−0.697	−1.342	1.042	1.307	1.763
19	If I were laid off, the organization would prefer to hire someone new rather than take me back.	−0.034	0.739	0.978	−1.036	−0.730	−1.461	1.402	1.220	2.009
6	The organization would ignore any complaint from me.	0.114	0.734	1.454	−1.555	−1.313	−1.356	0.241	0.748	1.652
28	The organization cares more about making a profit than about me.	−0.114	0.691	0.920	−0.927	−1.234	−1.489	1.358	1.266	1.948
15	The organization feels there is little to be gained by employing me for the rest of my career.	0.012	0.687	1.116	−1.750	−1.679	−2.086	0.686	0.819	1.529
12	If the organization found a more efficient way to get my job done they would replace me.	−0.063	0.678	0.955	−1.630	−1.340	−1.822	0.919	0.967	1.748
2	If the organization could hire someone to replace me at a lower salary it would do so.	−0.060	0.650	0.755	−1.312	−1.107	−1.566	0.921	1.040	1.898
31	The organization feels that anyone could perform my job as well as I do.	−0.129	0.625	0.718	−1.578	−1.515	−1.725	1.277	1.357	1.833
34	If my job were eliminated, the organization would prefer to lay me off rather than transfer me to a new job.	−0.018	0.623	0.769	−1.191	−1.330	−2.490	1.202	0.977	1.169
11	The organization would fail to understand my absence due to a personal problem.	−0.046	0.527	0.644	−1.541	−1.807	−2.332	0.737	0.856	1.512
	Interfactor correlation Factor 2	−0.014
	Eigenvalue	11.50	9.52

Abbreviation: SPOS-J, Japanese version of the Survey of Perceived Organizational Support. The bold text indicates which of the two factors each item belongs to as a result of the EFA.

Reliability (internal consistency and test–retest reliability)

Table 3 shows the Cronbach a coefficients, Pearson correlation coefficients, ICCs, and SEM for SPOS-J (SPA) and SPOS-J (MNA). Cronbach a coefficients for SPOS-J (SPA) and SPOS-J (MNA) showed high internal consistency (.92 and .84, respectively). In addition, Pearson correlation coefficients and ICCs showed moderate test–retest reliability for SPOS-J (SPA) and SPOS-J (MNA) in the 2-week interval test–retest.

Table 3.

Items, structures, means, Cronbach α coefficients, and test–retest reliability of the final version of the scale (SPOS-J).

Confirmatory factor analysis					Factor characteristics a
Factor	Name	Item no.	Loading	Item in English	Min	Max	Mean	SD	Cronbach α	r	ICC (95% CI)	SEM (95% CI)
Factor 1	Sufficiency of positive aspects (SPA)	10	0.888	The organization is willing to extend itself in order to help me perform my job to the best of my ability.	0	6	2.70	1.21	.92	0.72 (P < .01)	0.72 (0.67-0.76)	0.80 (0.54-0.64)
		25	0.874	The organization cares about my opinions.
		9	0.825	The organization really cares about my well-being.
		21	0.811	The organization cares about my general satisfaction at work.
Factor 2	Minimality of negative aspects (MNA)	17	0.293	Even if I did the best job possible, the organization fails to notice.	0	6	3.19	1.21	.88	0.56 (P < .01)	0.55 (0.49-0.61)	0.98 (0.91-1.05)
		23	0.274	The organization shows very little concern for me.
		7	0.255	The organization disregards my best interests when it makes decisions that affect me.
		3	0.262	The organization fails to appreciate any extra effort from me.

Abbreviations: ICC, intraclass correlation coefficient; r, Pearson correlation coefficient; SEM, standard error of measurement; SPOS-J, Japanese version of the Survey of Perceived Organizational Support.

^aMean, SD, Min, Max, and Cronbach α were calculated using all responses in the baseline survey.

Convergent validity

Table 4 shows the correlations between the 2 subscales of the final version of the SPOS-J (8 items), SPOS-J (SPA) and SPOS-J (MNA), and all study variables, including the subscales of the 33-item version of the SPOS-J, antecedents (supervisor support, organizational justice [procedural and interactional], and role conflict) and outcome factors (job satisfaction, work engagement, organizational commitment, and turnover intention). The 2 subscales showed a small negative correlation with role conflict, a small to medium negative correlation with turnover intention, and small to large positive correlations with other antecedents and outcome factors.

Table 4.

Means, SDs, and correlations of variables used in the study (n = 6220).^a

	Mean	SD	SPOS-J 8-item subscales b
			(SPA)	(MNA)
1. SPOS-J 8 items
1-1 SPOS-J 8 items (SPA)	2.70	1.21	(.92)
1-2 SPOS-J 8 items (MNA)	3.19	1.21	.14*	(.88)
2. SPOS-J 33 items
2-1 SPOS-J 33 items (SPA)	2.72	1.07	.95*	.07*
2-2 SPOS-J 33 items (MNA)	3.18	1.05	.02	.93*
3. Supervisor support	2.68	0.81	.45*	.30*
4. Organizational justice (procedural)	3.13	0.81	.47*	.39*
5. Organizational justice (interactional/relational)	3.23	0.91	.52*	.37*
6. Role conflict	2.34	0.85	−.21*	−.13*
7. Job satisfaction	2.54	0.83	.42*	.27*
8. Work engagement	2.28	1.61	.43*	.04*
9. Organizational commitment	2.80	0.83	.46*	.19*
10. Turnover intention	2.69	1.38	−.35*	−.22*

Abbreviations: MNA, Minimality of Negative Aspects; SPA, Sufficiency of Positive Aspects; SPOS-J, Japanese version of the Survey of Perceived Organizational Support.

^aThe SPOS-J (8 items) is the final version, and the SPOS-J (33 items) is the version before the number of items was reduced using item response theory analysis.

^b*P < .01; Cronbach α coefficients are displayed in parentheses.

Known-group validity

Table 5 shows a comparison of the means of the SPOS-J subscales by demographic characteristics. The results of the 1-way ANOVA indicated significant differences between groups in the SPOS-J subscale scores based on age and organizational size. Specifically, SPOS-J (SPA) scores increased with age, and when limited to organizations with up to 500 employees, smaller organizations exhibited higher SPOS-J (SPA) scores. In contrast, SPOS-J (MNA) showed a decreasing trend with age up to the 50s, and no consistent trend was observed with organizational size changes. Tukey post hoc multiple comparisons revealed that for SPOS-J (SPA), significant differences were observed between the 60-64 age group and all other age groups, as well as between the 100-299 employees group and the 1000 or more employees group. Additionally, SPOS-J (MNA) demonstrated significant differences between the 20-29 age group and all age groups 40 years and older, and significant differences were found between the 1-99 employees group and the 500-999 employees group.

Table 5.

Known-groups validity of the SPOS-J subscales (n = 6220) (1-way ANOVA and Tukey post hoc multiple comparisons).

	SPOS-J subscales a
	SPA		MNA
	Mean (SD)	P value	Mean (SD)	P value
Age, y		<.001		<.001
20-29 (n = 1131)	2.60a (1.28)		3.36a (1.24)
30-39 (n = 1281)	2.64ab (1.25)		3.28ab (1.24)
40-49 (n = 1684)	2.67ab (1.18)		3.15c (1.22)
50-59 (n = 1455)	2.76b (1.18)		3.05c (1.20)
60-64 (n = 669)	2.97c (1.09)		3.18bc (1.09)
Organizational size		.011		.018
1-99 employees (n = 2382)	2.70 (1.20)		3.26a (1.22)
100-299 employees (n = 892)	2.64a (1.23)		3.14 (1.22)
300-499 employees (n = 371)	2.62 (1.19)		3.21 (1.22)
500-999 employees (n = 492)	2.63 (1.17)		3.15 (1.21)
>=1000 employees (n = 1810)	2.78b (1.23)		3.14b (1.22)
Public offices (n = 273)	2.75 (1.18)		3.18 (1.11)

Abbreviations: ANOVA, analysis of variance; MNA, Minimality of Negative Aspects; SPA, Sufficiency of Positive Aspects; SPOS-J, Japanese version of the Survey of Perceived Organizational Support.

^aSuperscripts a, b, c indicate statistically significant differences between the groups. Groups without superscripts do not have significant differences with any other groups. Statistically significant difference: P < .05.

Interpretability

As shown in Table 3, the SDs for SPOS-J (SPA) and SPOS-J (MNA) at baseline were both 1.21, and the MCID for SPOS-J (SPA) and SPOS-J (MNA) was 0.6 points each.

Discussion

Reliability and validity

In this study, we developed the SPOS-J and evaluated its reliability and validity. Internal consistency was assessed using Cronbach a, and test–retest reliability was evaluated using Pearson correlation coefficient, ICC, and SEM. Factorial validity was examined by EFA, CFA, and IRT analyses; convergent validity by correlation analysis; known-group validity by ANOVA; and interpretability by determining MCID.

Regarding factorial validity, the 2-factor model comprising the SPOS-J (SPA) and SPOS-J (MNA) subscales showed a higher fit than the 1-factor model of the original version. Reduction of the item count to 8 through IRT analysis yielded an acceptable fit for this 2-factor model. These subscales consisted of items that queried positive content for SPOS-J (SPA) and negative content for SPOS-J (MNA). In developing an affective scale, like the original SPOS, it is common practice to include an equal number of positive and negative questions to reduce response set bias.^43,44 Nevertheless, reports indicate that using positive and negative items to form a unidimensional scale often leads to a 2-factor model, where negative items form a distinct scale from positive ones.^43–45 Knight et al⁴⁶ also found that items with positive and negative phrasing loaded on different factors. The distinct categorization of the SPOS-J subscale into positive (SPOS-J [SPA]) and negative (SPOS-J [MNA]) items in this study’s results may be attributed to the finding that merging both types of items into a unidimensional scale compromises its unidimensionality. Furthermore, unlike the original US version, which maintained unidimensionality, the loss of unidimensionality in the SPOS-J for Japanese participants could be attributed to differences in how Westerners and Japanese respond to positive and negative items.⁴⁷ Alternatively, considering the original SPOS was developed more than 35 years ago, its unidimensionality might be questioned due to evolving perceptions of organizational support. However, Worley et al,⁴⁸ in 2009, affirmed the SPOS as a unidimensional measure, indicating its consistent dimensionality over time. Thus, it seems improbable that historical changes in perceptions of organizational support have altered the unidimensionality of the original version. Instead, the presence of both positive and negative items in the unidimensional scale, combined with differences in response tendencies to these items between Japanese and Western individuals, likely explains the SPOS-J’s alignment with a 2-factor model in this study.

In this validation, conducting CFA on the 33 items of the SPOS-J showed that the model fit was not necessarily sufficient even for the 2-factor model. One reason for this is that these items may include questions that are inappropriate or meaningless depending on the country’s culture, employment system, or specific industries (eg, civil servants and other industries with high employment security). In fact, when examining the reliability and validity of the SPOS for teachers, Dolma and Torun¹⁸ excluded items from the 36 items of the SPOS that were not appropriate for their employment environment as teachers. The original version of the SPOS¹ by Eisenberger et al was basically developed to understand POS for US workers. Although this study aimed to decrease the number of items through IRT analysis, reducing items by considering cultural contexts, particularly to better match the scale with Japanese cultural and labor practices, is another viable approach.

For reliability, the internal consistency of SPOS-J, SPOS-J (SPA), and SPOS-J (MNA) was sufficient with Cronbach a coefficients, and Pearson correlation coefficients and ICCs showed moderate test–retest reliability for SPOS-J (SPA) and SPOS-J (MNA). To the best of our knowledge, we could not find any studies that examined test–retest reliability abroad, but this study suggests that, at least for the SPOS-J, it has a certain level of test–retest reliability.

Regarding convergent validity, this study examined the relationship between supervisor support, organizational justice (procedural and interactional), and role conflict as antecedents of the SPOS-J. We also examined the relationship between job satisfaction, work engagement, organizational (affective) commitment, and turnover intention as outcome factors on the SPOS-J. The patterns of positive and negative correlations between the 2 SPOS-J subscales, SPOS-J (SPA) and SPOS-J (MNA), and their expected antecedents and outcome factors matched those of the original version of the SPOS. The SPOS-J (SPA) and SPOS-J (MNA) demonstrated small to medium correlations with supervisor support. Several studies examining supervisor support as a variable affecting POS have confirmed a highly positive relationship between supervisor support and POS,^3,4,49 which is consistent with the results of this study. The discovery of a moderate positive association between the 2 SPOS-J subscales and both procedural and interactional justice within organizational justice aligns with previous studies.^2,15,50 Contrary to positive factors like supervisor support and organizational justice, stressors constitute another antecedent to POS. However, research on POS has focused primarily on the effects of stressors related to their role in the organization. Regarding role conflict, a role-related stressor, several studies^8,51 have shown that role conflict and POS are negatively correlated, but the strength of the correlation ranges (r = −0.20 to −0.64). The results of this study showed a small negative correlation (r = −0.20, −0.13), which was close to the results of Hutchison et al.⁵¹ Job satisfaction has been empirically found to have a medium positive correlation with POS¹⁰ and was generally consistent with the results of this study (r = 0.40, 0.24). Work engagement has also been demonstrated to have a positive correlation with POS,⁹ aligning with the findings of this study (r = 0.42, 0.03). Based on social exchange theory and the idea of reciprocity norms, Eisenberger et al¹ argue that POS can foster a sense of obligation to the organization and increase affective commitment to the organization. The results of this study also showed that the SPOS-J (SPA) and SPOS-J (MNA) had middle and small positive correlation (r = 0.43, 0.18) with affective commitment, and many existing studies^5,11,52 found a positive relationship between POS and affective commitment, although the strength of the correlation varied (r = 0.38 to 0.81). This variation may have been caused by differences in the target population. In a meta-analysis by Rhoades and Eisenberger,¹⁵ among withdrawal behaviors (turnover intentions, turnover, and other withdrawal behaviors), the strongest relationship with POS was found for turnover intentions, and turnover intentions have been empirically shown to have a middle negative correlation with POS.¹⁴ This result was consistent with the results of this study (r = −0.32, −0.21).

In this study, the SPOS-J (SPA) showed higher correlation coefficients with all relevant factors compared with the SPOS-J (MNA). This difference might stem from the negative wording of the SPOS-J (MNA) items. Benson and Hocevar⁴⁴ observed that including both positive and negative items in a questionnaire could lower the validity of negatively worded items. Eisenberger et al,⁵³ developers of the original SPOS, suggested employing 10 items with exclusively positive wording to circumvent issues like poor internal consistency and a multifactorial structure.

The results of the 1-way ANOVA revealed significant differences in the mean scores of each SPOS-J subscale across groups based on age and organizational size. Among the subscales, SPOS-J (SPA) increased with age and exhibited a U-shaped curve concerning organizational size, with the lowest scores in the group of 300-499 employees. This result aligned with our hypothesis. On the other hand, SPOS-J (MNA) tended to decrease with age up to the 50s, and no consistent trend was observed with changes in organizational size. This discrepancy may be due to the lower validity caused by the negative wording of the SPOS-J (MNA) items,⁴⁴ indicating that further investigation into the validity of SPOS-J (MNA) is necessary. Our study results support the known-group validity of the SPOS-J subscales, at least for SPOS-J (SPA), indicating that SPOS-J (SPA) can distinguish between different groups as hypothesized.

Regarding interpretability, this study used 0.5 SD of the baseline score as the MCID, resulting in an MCID of 0.6 points for both SPA and MNA. Therefore, a change of 0.6 points in SPA and MNA is considered practically significant. Regarding the SD method, a range of 0.4 to 0.5 SD is considered clinically significant, and previous studies have shown that most MCIDs fall within 0.5 SD.^42,54,55 Moreover, a review by Ousmen et al⁵⁶ reported that the most frequently cited distribution-based approach is 0.5 SD of the baseline score. Therefore, this study used 0.5 SD of the baseline score as the MCID. The distribution-based approach has the advantage of being simple to use because it does not require an external criterion. However, this approach yields the same MCID results for both deterioration and improvement of the measure. Although this simplifies interpretation, it is often questioned because MCIDs for improvement and deterioration can differ.⁵⁷ Therefore, it is necessary to set external criteria in the future to calculate MCID more accurately.

Practical implications

In recent years, there has been a great deal of research on the relationship between work engagement, which has received attention as a predictor of physical and mental health⁵⁸ and job performance,⁵⁹ and POS, and work engagement has been found to be positively related to POS.⁹ This study also found that the SPOS-J (SPA), a subscale of the SPOS-J, had a moderate positive correlation with work engagement. The concept of strategically working to improve POS by showing active consideration for employee well-being, and aiming to improve corporate performance and curb employee turnover by improving employee job satisfaction and work engagement, is highly compatible with the concept of health and productivity management⁶⁰ that is gaining popularity in Japan. Therefore, making SPOS-J available as an evaluation indicator for organizations may contribute to the further development of occupational health practice and health and productivity management in Japan through the provision of a new evaluation indicator. Furthermore, this study revealed that small and medium-sized enterprises with 300 to 499 employees and younger employees tend to have relatively low levels of POS. It is imperative for companies to engage in proactive strategies aimed at elevating POS among these targeted worker groups. Additionally, the results of this study showed that the SPOS-J (SPA), comprising 4 items, demonstrated sufficient internal consistency and supported known-group validity. Its correlations with other related factors were comparable to those of the original SPOS. Therefore, from a practical standpoint, it may be feasible to use only the 4-item SPOS-J (SPA) as a shortened version of the SPOS-J to reduce the burden on respondents. In fact, Eisenberger et al,⁵³ the developers of SPOS, have also suggested using only positively worded items to avoid issues such as lack of internal consistency and multifactor structures. Moreover, in the United States, where the original SPOS was developed, there is a study that used a shortened version of the SPOS consisting of only the 4 positive items.⁶¹

Limitations

This study has several limitations. First, this study is an internet-based survey and may not represent the general population of Japanese workers, because those who do not have internet access or are not registered with an online survey firm are omitted from the target population. However, because this study is based on data from a labor force survey conducted by a public agency, and the sample was collected stratified by gender, age, and region of residence to reduce sampling bias, the generalizability of the study’s results is considered sufficient. Second, this study is based on survey data from self-reports, and the present results should be replicated in the future using objective measures (eg, measures related to performance, leaving a job, and other behaviors). Third, this study did not assess criterion validity, cross-cultural validity, or responsiveness.

Conclusion

In conclusion, this study confirms that, among the 8-item SPOS-J, at least the subscale SPOS-J (SPA) has supported reliability and validity, and can be used in a Japanese context.

Author contributions

K.O., T.N., H.E., K. Mafune, and K. Mori designed the study; K.O., T.N., H.E., and K. Mori participated in the development of the Japanese version of the scale; K.O., A.I., and K. Mori analyzed the data; K.O. led the writing of the manuscript, and all other authors were involved in the drafting the manuscript. All authors read and approved the final manuscript.

Funding

This study has been funded by the Japan Society for the Promotion of Science (JSPS) KAKENHI (JP21K21155), and the HASEKO Corporation.

Conflicts of interest

None declared.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.