CC BY-NC-ND
0Available online at www.sciencedirect.com
SciVerse ScienceDirect
Procedía - Social and Behavioral Sciences 65 (2012) 1013 - 1019
International Congress on Interdisciplinary Business and Social Science 2012
(ICIBSoS 2012)
Personality Traits, Work-Family Conflict and Job Satisfaction: Items Validity using Rasch Measurement
Approach
Nurhazirah Hashima, Nor Irvoni Mohd. Isharb, Wan Edura Wan Rashid & Mohd.
Saidfudin Masodid
abcFaculty of Business Management, Universiti Teknologi MARA Malaysia Accounting Research Institute, Universiti Teknologi MARA Malaysia nurhazirah@puncakalam. uitm. edu. my
Abstract
The role of women has changed dramatically in the present scenario. The claim has been supported by a general consensus among researchers that many married working women experience significant work-family conflict (WFC). However, the personality traits (PT) have been only minimally considered in the studies of WFC especially in Malaysia. To fill in this gap, a study was conducted to investigate the influence of PT on the relationship between WFC and job satisfaction (JS) among married female teachers at selected schools in Klang Valley. Hence, the aim of this paper is to look at how the Rasch Model is able to facilitate the revalidation process of this instrument construct validity. A number of 38 items were used to measure the personality traits of 90 married female teachers and was tested using the Rasch Measurement Model Analysis. Evaluation of the mean square infit and outfit suggests that the data exhibited; fit the model, which means they are likely to be measuring the single dimension intended by the construct theory. In addition, responses to the statements in the questionnaire show greater consistency and higher reliability coefficient. The result of this paper also confirms the robustness and the validity of the instrument that can be applied in any organizations in Malaysia.
© 2012 Published by 111sevier Ltd.
Selection and/or peer-review under responsibility of JIBES University, Jakarta
Keywords: Personality traits, Work-Family conflict, Job satisfaction, Women, Malaysia, Rasch Measurement Model
1877-0428 © 2012 Published by Elsevier Ltd.
Selection and/or peer-review under responsibility of JIBES University, Jakarta doi:10.1016/j.sbspro.2012.11.235
1. Introduction
Recently, the issue of work-family conflict (WFC) has become a growing topic of interest among researchers due to the emerging roles of work and family which gives implications to both organization and employees. Although, stress researchers have given special attention to the role of personality factors in stressors-strain relationship where work-family conflict is considered as a potential stressor that leads to various form of stress reactions (McShane & Von Glinow, 2010), there is still lack of study that explore personality traits in this context (Nurhazirah & Abdul Kadir, 2011). Despite the fact that most existing research on this area was aggressively conducted in the Western societies, the understanding on the effects of work-famly conflict on women's job satisfaction is critically important in Asian countries where women have recently made remarkable advancement in the business world (Nurhazirah & Abdul Kadir, 2011).
To date, the personality traits (PT) factors have been only minimally considered in studies of WFC especially in Malaysia (Nurhazirah & Abdul Kadir, 2011). The lack of empirical evidence research creates a gap between theory and practices of personality traits among the married female employee in Malaysia. Thus, this paper aims to examine the extent of relevant items on PT, WFC and job satisfaction (JS) fits the expectations of the Rasch Measurement Model. In this study, the data is analyzed using Rasch Measurement Model which provides a linear ruler scale for intangible measurement (Bond & Fox, 2007).
2. Literature Review
Researchers has demonstrated that WFC is bi-directionally which means that the conflict arising when work roles interfere with family roles (WIF) and family related roles interfere with work related roles (FIW) (Kinnunen & Mauno, 1998). Consistent with the previous claim, Hammer (2003) noted that WIF and FIW have different antecedents and consequences. Antecedents of WIF include long hours, lack of supervision, and other work role stressors and characteristics which the consequences have been related to life satisfaction (Adams, King, & King, 1996), family satisfaction (Beutell & Witting--Berman, 1999), alcohol abuse, depression, and poor physical health (Mc Shane, 2010).
Furthermore, more research has explored the effects of WFC to the women's job satisfaction which reported that work-family conflict was negatively related to married women's job satisfaction (Aryee, 1993; Kinnunen & Mauno, 1998). Most of the existing research on the relationship between work and family conflict has been conducted in Western societies, but as more women in non-Western societies join the work force, understanding the effects of WFC on these women's satisfaction has become increasingly important (Aminah, 1996). There are also studies that found personality traits act as moderators on the relationship between work-family conflicts and well-being outcomes (Kinnunen, 2003).
Personality can be defined as the set of unseen characteristics and processes which reflects a stable pattern of behavior in response to ideas, objects, or people in the environment (Daft, 2008). Personality was approached according to the framework of the Big Five which consist of conscientiousness, extraversion, agreeableness, emotional stability and openness to experience (Goldberg, 1992). These personality traits are expected to play some important roles in influencing the relationship between work-family conflict (WFC) and job satisfaction.
3. Methodology
A set of 38-items questionnaire using a five-point likert rating of between 1 to 5, where; 1 - Strongly Disagree, 2 - Disagree, 3 - Fair, 4 - Agree, 5 - Strongly Agree were distributed to 90 married female teachers at selected primary and private schools in Klang Valley, Malaysia. Three (3) dimensions is under
scrutiny i.e Personal Traits (PT), Work Family Conflict (WFC) and Job Satisfaction (JS). Data collected were then tabulated and analyzed using Rasch Measurement Model, a rating scale model that uses the item response theory was applied to test the overall data fit (Bond & Fox, 2007) with the aid of Rasch Analysis Software - WinSteps® version 3.72.1.
4. Findings
Three (3) tests were performed on the data. A total of 3420 data points were analyzed with a yield of log likelihood Chi-Square of 7824.28 with 3290 degree of freedom (d.f.) at p=0.0000.
4.1 Reliability
The Summary Statistics in Table 1 revealed that the Cronbach-a value was 0.68 which is acceptable for exploratory purposes indicating the test reliability in measuring the interaction between the respondents and the items in the respective dimensions viz; PT, WFC and JS. It further shows that the reliability of item difficulty estimates is high at 0.98 indicating sufficient number of items to measure what it needs to measure (Linacre, 2003). The item separation index of 6.34 indicates that the items can be separated into six (6) difficulty strata. As item reliability indicates the ability of the test to produce the hierarchy of items along the measured variables (Linacre, 2003; Alderson, Clapham & Wall, 1995) a reliability coefficient of 0.98 suggests that this order of item hierarchy will be replicated with a high degree of probability if the items were given to other comparable cohorts (Nor Irvoni & Saidfudin, 2012). However, the power of discrimination i.e. the ability of the items in the instrument to separate the Persons is only a mere 1.27 Person Separation. A small Person Mean of 0.07, [SE 0.19]logit indicates generally a weak agreement to family conflict due to women at work.
The Person Measures reliability coefficient is considerably low at 0.62 which indicates that this order of Person hierarchy will less likely be replicated if they were given another set of items of the same construct (Azrilah, 2011). This is attributed to the considerable misfitting Person responses in the data. If responses to the statements in the questionnaire on the other hand showed greater consistency, then this shows a higher reliability coefficient of the responses data.
Table 1. Summary Statistics
CRONBACH ALPHA (KR-20) Person RAW SCORE RELIABILITY = .68
Items: 38 MEASURED Person 90 MEASURED
total SCORE COUNT MEASURE MODEL ERROR INFIT MNSQ ZSTD OUTFIT | MNSQ ZSTD | TOTAL SCORE COUNT MEASURE MODEL ERROR INFIT MNSQ ZSTD OUTFIT | MNSQ ZSTD |
MEAN 2 80.8 90.0 S.D. 60.0 .0 MAX. 3 76.0 90.0 MIN. 196.0 90.0 .00 .86 1.17 -1.66 .13 .02 .17 . 11 .99 -.2 .24 1.7 1.60 2.7 .50 -3.1 1.00 -.1 | .24 1.6 | 1.67 3.1 | .54 -2.8 | | MEAN | S.D. | MAX. | MIN. 118.5 9.4 148.0 95.0 38.0 .0 38.0 38.0 .07 .35 1.23 -.78 .19 .00 .22 .19 1.00 -.6 .74 2.8 5.06 9.9 .21 -5.5 1.00 .82 6.10 .18 -.6 | 2.8 | 9.9 | -5.4 |
REAL RMSE .13 TRUE SD S.E. OF Item MEAN = .14 .85 SEPARATION 6 34 Item RELIABILITY .98 | | REAL | S.E. RMSE .21 TRUE SD OF Person MEAN = .04 .27 SEPARATION 1.27 Person RELIABILITY ------ | .62 | ------- |
Item RAW SCORE-TO-MEASURE CORRELATION 3420 DATA POINTS. LOG-LIKELIHOOD CHI- = -.99 SQUARE: 7824 .28 with 3290 d.f. p=0.0000
Next, look at the length of the ruler. These measures gave the Item Ruler length of 2.83 logits against the Person Ruler of 2.01logit longer by 0.82logits to measure the Person. This gives an overall overview that the item difficulty is spread over 2.83logit whilst the Person ability measured is shorter. However, take note on the top of the Person ruler we see Person Free Items where the PersonMax+1.23logit as against ItemMax+1.17logit hence no corresponding items for these respondents. The opposite observation is seen at the bottom where Item Free Person is found instead; lowest Item is measured at -1.66logit against the lowest Person measured higher at -0.78logit. These items are classified as extremely easy
hence overly agreeable items which does not discriminate the Persons at all. Generally these items are likely candidates to be dropped.
4.2 Fit Statistics
The study refers to the common logit scale, since this is the same scale that is used in measuring both the Person ability and the Item difficulty; comparing both variables on the same interval scale. Rasch requires further verification by looking at the OUTFIT column for Mean Square value; INFIT MNSQ = y, 0.5 < y < 1.5. Six items; Item 2-WFC2, 25-PT7, 24-PT6, 13-JS3, 14-JS4 and 11-JS1 were found beyond this parameter. However, if an item having larger MNSQ than the sum of the mean of IMNSQ and SD, that gives an indication of possible high z-std; in this case 1.23logit thus item misfit. Table 1 shows four items; Item 14-JS4, 2-WFC2, 25-PT7 and 24-PT6, is a misfit with MNSQ > 1.23logit and further checks was done on the z-Std value, where z-Std = z, -2 < z < +2; confirmed that the aforesaid four (4) items viz; Item 14-JS4, 2-WFC2, 25-PT7 and 24-PT6 falls outside the set range.
Table 2. Item Measure Order
| ENTRY TOTAL TOTAL MODEL INFIT OUTFIT PT-MEASURE EXACT MATCH!
|NUMBER SCORE COUNT MEASURE S.E. MNSQ ZSTD MNSQ ZSTD CORR. EXP. OBS8 EXPS! Item !
36 196 90 1 17 .12 96 - 2 92 _ 5 .05 .29 58 9 51 5! PT18 !
8 198 90 1 14 .12 1 16 1 1 1 09 6 .59 .30 58 9 51 4! WFC8 !
7 202 90 1 08 .12 1 20 1 3 1 13 9 .53 .30 52 2 50 4! WFC7 !
9 203 90 1 07 .12 1 06 4 1 02 2 .50 .30 57 8 50 3! WFC9 !
6 205 90 1 04 .12 97 - 2 94 - 4 .56 .30 53 3 49 4! WFC6 !
4 208 90 1 00 .12 1 08 6 1 07 5 .46 .30 54 4 48 8! WFC4 !
10 211 90 95 .12 1 09 7 1 09 7 .50 .30 52 2 46 9! WFC10!
2 214 90 91 .12 1 29 2 1 1 27 1 9 .44 .30 42 2 45 0! WFC2 !
38 214 90 91 .12 69 -2 5 68 -2 5 .20 .30 55 6 45 0! PT2 0 !
1 215 90 90 .12 1 21 1 5 1 19 1 4 .45 .31 41 1 44 0! WFC1 !
5 217 90 87 .12 1 03 2 1 01 2 .51 .31 46 7 43 8! WFC5 !
26 220 90 83 .12 92 - 6 91 - 7 .31 .31 48 9 42 0! PT8 !
3 225 90 77 .11 1 02 2 1 01 1 .49 .31 41 1 39 9! WFC3 !
35 247 90 49 .11 87 -1 2 90 - 8 .17 .31 36 7 30 5! PT17 !
25 248 90 48 .11 1 33 2 7 1 39 3 1 .03 .31 26 7 30 5! PT7 !
32 248 90 48 .11 74 -2 5 74 -2 4 .34 .31 47 8 30 5! PT14 !
24 274 90 16 .11 1 27 2 3 1 33 2 7 -.07 .30 26 7 26 9! PT6 !
22 275 90 15 .11 83 -1 6 84 -1 4 .28 .30 41 1 26 9! PT4 !
16 287 90 00 .11 1 31 2 5 1 35 2 7 .03 .29 22 2 28 5! JS6 !
29 293 90 - 07 .11 81 -1 7 86 -1 2 .36 .29 35 6 28 9! PT11 !
33 298 90 - 14 .12 70 -2 6 70 -2 6 .30 .29 42 2 30 9! PT15 !
19 307 90 - 26 .12 67 -2 8 72 -2 2 .27 .28 47 8 36 2! PT1 !
17 313 90 - 35 .12 1 02 2 1 04 3 .23 .27 35 6 40 0! JS7 !
23 315 90 - 38 .12 83 -1 2 88 - 8 .19 .27 38 9 42 5! PT5 !
34 317 90 - 41 .12 76 -1 8 74 -1 9 .41 .27 48 9 45 8! PT16 !
28 318 90 - 42 .12 99 0 1 06 4 .21 .27 48 9 46 0! PT10 !
27 321 90 - 47 .12 75 -1 8 77 -1 6 .18 .26 60 0 48 3! PT9 !
37 331 90 - 63 .13 78 -1 4 77 -1 4 .19 .25 58 9 55 5! PT19 !
30 335 90 - 70 .13 56 -3 1 57 -2 8 .19 .25 65 6 59 4! PT12 !
18 342 90 - 83 .14 95 - 2 99 0 .25 .24 66 7 62 4! JS8 !
21 342 90 - 83 .14 83 - 9 89 - 5 .17 .24 64 4 62 4! PT3 !
13 343 90 - 85 .14 1 33 1 7 1 35 1 8 .15 .24 61 1 63 0! JS3 !
15 354 90 -1 08 .15 1 05 3 1 08 5 .17 .23 68 9 64 5! JS5 !
14 355 90 -1 10 .15 1 60 2 7 1 67 2 9 -.02 .22 62 2 65 0! JS4 !
20 366 90 -1 38 .16 50 -3 0 54 -2 7 .22 .21 77 8 65 6! PT2 !
12 367 90 -1 40 .16 1 14 7 1 16 8 .10 .21 67 8 65 5! JS2 !
11 369 90 -1 46 .17 1 24 1 2 1 26 1 3 .05 .21 65 6 65 2! JS1 !
31 376 90 -1 66 .17 1 12 7 1 07 4 .18 .20 63 3 63 9! PT13 !
| MEAN 280.8 90.0 00 .13 99 _ 2 1 00 _ 1 51 2 47 2!
| S.D. 60.0 .0 86 .02 24 1 7 24 1 6 12 8 12 4!
SUM OF MEAN + S.D:1.23
Scrutiny of items from the same dimension having the same measure shows only Item 25: PT7-"low agreeableness" and Item 32: PT14 -"low agreeableness"; +0.481ogit, [SE:0.11]. Respondents see the item as measuring the same thing. Item whose MNSQ is nearer to 1 and z-Std nearer to 0 is deemed a better fit
thus Item 32 [MNSQ:0.74 ; Zstd: -2.5] is maintained and Item 25[MNSQ:1.33 ; Zstd: 2.7] should be deleted or rephrased to preserved content validity. Good Item Construct shall be like "PT Items" in the vertical direction. Items in horizontal direction shows redundancy of items in the same construct measuring the same thing thus poor construct. It is not desirable.
Next, the quality of items is determined by checking the attributes Point Measure Correlation (PMC). Generally all the items; except for Item 14-JS4 and Item 24-PT6, have positive Point Measure Correlation (PMC). PMC refers to the items whether it is measuring in the correct direction. A good item shall generally be with positive PMC and stand in the range of 0.28 - 0.86. The narrower the PMC the better precision the item has in measuring the said variable. An item with negative PMC is likely to be dropped. Table 2 shows that two (2) item 14-JS4 and 24-PT6 register negative PMC which is undesirable where items are measuring in the wrong direction.
The very strength of Rasch Measurement Model is the precision of measurement where each item has its own measurement error reflected by the 'Standard Error', S.E.. Though the items may have a good measurement quality but if the S.E. for each item measurement is larger than the measure separation, the clarity of measurement becomes cluttered up. Let us take Item 9-WFC9; measured +1.07logit, S.E. +/-0.12. On the upper limit this item reads as +1.19logit. This changes the position drastically as being the most difficult item above Item36-PT8; +1.17logit. Simultaneously this item is also seen as +0.95logit level of difficulty which is three (3) tier easier than average.
Now the actual position of Item 9-WFC9 lies anywhere between the six (6) tiers and it becomes more clouded with the other two (2) items above and three(3) item below. All these WFC items 8,7,9,6,4 and 10 need to be reviewed to improve clarity of measurement by achieving the S.E. smaller than the measure separation. Otherwise, the item may be dropped if no plausible explanation can be offered for such response pattern. Similar pattern is seen for item PT5,16,10 and 9. Thus the Person Separation registers a low index of 1.27. Improving these items will result in higher Person Separation.
4.3 Persons - Items Distribution Map (Wright Map)
This is the heart of Rasch Analysis (see Fig. 1). Similar to the traditional histogram tabulation, it allows both the person and item to be mapped together but now on the same logit scale giving a clearer view of the persons' ability and the relevant item difficulty for a better understanding of how a person's behavior correlates to each respective items (Nor Irvoni & Saidfudin, 2012). It shows all and more importantly the logical hierarchy of difficulty based on the conceptual theory put under test. This will be the premise of the instrument construct validity acceptance. When the item difficulty hierarchy is in place then it is said the instrument has the construct validity. All the items have a fair hierarchical order with an item measuring range of large 3.96logit but require dire review of items in WFC dimension where the items are very much lack clarity in measurement.
From the map it is evident that a large number of items of each dimension can be found along the continuum on which the majority of teacher's abilities fall. However, there are easy items that are Person free at the bottom of the map marked G1. This means some of the questions were too easy for the teachers. G-1s are Person Free Items but of different psychometry which means item homogeneity exists despite of the differences in the teachers' perception (Nor Irvoni & Saidfudin, 2012). Hence, it can be concluded that items that belonged to the G-1 category are off target items that is extremely easy to endorse where all respondents tend to strongly agree with the items. These items does not contribute much to the measurement, hence it should be dropped or rephrased so that it could better measure the teachers' ability. It cannot discriminate the Person distinctly.
Person - MAP - Item <more>|<rare>
<less> EACH "
<frequ> " IS 1.
PT2 0 PT8 PT14 PT17 PT7 PT4 PT6
JS7 PT1 PT16
PT10 PT9
PT12 PT19
JS3 JS8 PT3
JS2 PT2
JS1 PT13
WFC10 WFC4 WFC6 WFC7 WFC1 WFC2 WFC3 WFC5
Item mean = '0' set
Person mean = Ologit
G1: person free items.
Off target items — extremely easy to endorse Can be dropped / to be rephrased
PT_13 is the most agreeable item / easiest to agree with
########
Figure 1. Person — Item Map
5. Conclusion
Rasch offers a better method than pure frequency counts and percentage reporting as it transformed the ordinal data into measurable ratio - the highest level of measurement and put the data on a linear scale called the 'Logit Ruler'. As a result, responses to the statement in the questionnaires showed greater consistency and a higher reliability coefficient. Thus, it confirms the robustness and the validity of the instrument for application in any organization.
As for instrument refinement, some of the existing items (G1) need to be reviewed and rephrased to a more difficult item to provide a much more accurate and valid measurement for JS and PT construct. There is also the need to review WFC construct to represent the hierarchy of difficulty so that will give have a better spread vertically in measuring the teachers' level of agreement.
The findings could also serve as guideline in formulating and executing work-family policy in public and private sectors in Malaysia. The development and validation of the instrument in measuring personality traits , WFC and job satisfaction is an innovation to the Malaysian context. This new development is able provide a strong basis for a valid instrument construct that gives a better and true linear measure in measuring personality traits particularly in WFC issue and the related job satisfaction. This paper has indeed shown that Rasch Analysis facilitate the addition of new knowledge to existing literature related to work family conflict and job satisfaction study easier but yet in a more systemic way.
References
Adams, G. A., King, L. A. & King, D. W. (1996). Relationship of job and family involvement, family support, and work-family conflict with job and life satisfaction. Journal of Applied Psychology, 81, 411- 420.
Alderson, J. C., Clapham, C. and Wall, D. Language Test Construction and Evaluation. Cambridge: Cambridge University Press, 1995.
Aminah, A. (1996). Work-family conflict among married professional women in Malaysia. The Journal of Social Psychology, 136(5), 663-665.
Aryee, S. (1993). Dual-earner couples in Singapore: an examination of work and non-work sources of their experienced burnout. Human relations, 46, 1441-1468.
Azrilah, A.A. Rasch Model Fundamentals: Scale Construct and Measurement Structure. Kuala Lumpur : Azrilah, 2011.
Barrick and Mount, (1991). The Big Five personality dimensions and job performance: A meta-analysis. Personnel Psychology, pp 1-26.
Bond, T.G & Fox, C.M. Applying The Rasch Model, Fundamental Measurement in the Human Science. 2nd ed. New Jersey: Lawrence Erlbaum Publishers, 2007.
Daft, R. L. (2008). Leadership, 5th ed. Cincinnati, OH: Thompson/South-Western.
Goldberg, L. R. (1992). The development of markers for the Big-Five factor structure. Psychological Assessment, 4, 26-42.
Hammer, L. (2003). Work-family role conflict. Sloan work and family, Encyclopedia Entry, 1-3.
Kinnunen, U., & Mauno, S. (1998). Antecedents and outcomes of work-family conflict among employed married women and men in Finland. Human Relations, 51, 157-177.
Linacre, J. M. (2005) Measurement, Meaning and Morality (Electronic Version). Rasch Research Papers, Explorations & Explanations: Research papers and Memoranda. Retrieved 14 June from http://www.rasch.org/memo71.pdf
McShane and Von Glinow, (2010). Organizational Behavior, McGraw-Hill, 5th edition.
Nor Irvoni, M.I. & Mohd Saidfudin, M. (2012). Students' Perception towards Quality Library Service Using Rasch
Measurement Model. in Proceedings of International Conference on Innovation Management and Technology Research, May-21-22, 2012, Malacca, Malaysia,668-672.Malaysia: IEEE.
Nurhazirah, H. & Abdul Kadir, O. (2011). Moderating Effect of Personality Traits on the Relationship between Work-Family Conflict and Job Satisfaction. Unpublished conference paper. ASEAN Conference on Behavioural-Environment Studies, Bandung, Indonesia.
