January 18 2002 Dear Colleagues The Australian group is meeting shortly to discuss the form of the questionnaire for our third survey. Attached you will find a rather scrappy document that provides some statistics, derived from the Survey #2 data, on the performance of individual items. Any comments or advice will be appreciated. Discussion document in preparation for Survey #3 Bob Cummins and Julie Pallant Draft #3: 18/1/02 Table of Contents 1. Factor analysis from Survey #2 1.1 All personal well-being items (N=9) 1.2 All national well-being items (N=12) 1.3 Personal and National items together (N=21) 1.3.1 Three-factor solution 1.3.2 Two-factor solution 1.4 Conclusions and recommendations 2. Exploratory factor analyses for gender and event – Looking for a third factor 3. Varimax vs. Oblimin 4. All national items regressed against ‘How satisfied are you with life in Australia?’ 4.1 Standard regression 4.2 Stepwise regression 4.3 Hierarchical regression 4.4 Hierarchical x Gender 5. Index item intercorrelations 5.1 Personal index 5.2 National index 6. Personal Index item validation (from Survey #1) 6.1 Prediction of other variables by the index domains 7. Inter-item correlations 1. Factor analyses: Survey #2 (N = 3,000) 1.1 All personal well-being items (N=9) All items inter-correlated .19 (relationships vs. safety) to .63 (standard of living vs. life as a whole). All variables correlated >.3 with at least one other variable. KMO = .892 Bartlett chi-square = 6333.76, df = 36, p=.000 Two-factor solution (Principle Component Analysis; Oblimin with Kaiser normalisation, Pattern matrix). Factor 1 ‘Personal’ Happiness Life as a whole Relationships Achievements Standard of living Health .86 .82 .82 .71 .66 .37 Safe you feel Future security Comm. connectedness Eigenvalue Variance explained Total variance explained = 57.84 Note: Factor 2 ‘Social Capital’ .93 .58 .56 4.19 46.53 1.02 11.30 Here and throughout the document, items forming the ‘Index of Personal Wellbeing’ are in bold. The other non-bolded items are additional personal items included in Survey #2. Correlation between the factors = .455 Comment: The equivalent analysis from Survey #1 produced a single factor. The emergence of a separate ‘Security/Social Capital’ factor is new. However, The 7 items of the P-Index hang together as a single factor if left to their own devices, and the second factor is weak. There are further analyses pertaining to this putative new factor in 1.3. 2 1.2 National Well-Being (all items, N=12) All items inter-correlated .31 (health sciences vs. life in Australia) to .57 (family support vs. health services). KMO = .930 Bartlett chi-square = 9214.15, df = 66, p = .000 One-factor solution (Component matrix) Economic stability Business Economic situation How Australia is governed Wealth/income distribution Family support Social conditions State of environment Health services National security Life in Australia Trust in people Factor .75 .73 .73 .72 .72 .70 .70 .68 .66 .62 .60 .45 _________________________________________________ Eigen value Variance explained 5.50 45.81% _________________________________________________ Note: The ‘Index of National Wellbeing’ items are in bold. Comment: The analysis is failing to distinguish between ‘Index’, other national items, and social capital (Trust). While the six N-Index items do form a single factor by themselves, the above analysis does indicate the possibility that some items, (eg National security), could be replaced by others (eg Economic stability). 3 1.3 Personal and National items together (N=21) KMO = .933 Bartlett chi-square = 16,403.11, df = 210, p = .000 1.3.1 Three factor solution Factor 1 National WB How Australia is governed Family support Economic stability Wealth/income distribution Business Economic situation Health services Social conditions State of the environment Life in Australia National security Happiness Life as a whole Relationships Standard of living Achievements Future security Health How safe you feel Trust in people Comm. connectedness .73 .71 .71 .70 .69 .69 .68 .68 .67 .55 .53 Eigen values 7.11 Variance explained 33.85 Total variance explained = 51.85% Comment: Factor 2 Personal WB Factor 3 Social Capital .46 .83 .82 .73 .71 .70 .51 .47 .47 .43 .77 .60 .48 2.69 12.79 1.09 5.21 The third factor ‘social capital’ has captured National, Personal, and Social Capital items. However, it just fails to form an acceptable factor due to complex loadings, which only leaves two clean items. Could the addition of more items pertaining to ‘social capital’ yield a coherent third factor? Possibly not, because the three cross-loaded items had approximately equal degrees of shared variance with both factors, and in 2/3 cases the loading was higher on the National or Personal factor. 4 1.3.2 Personal and National items together (N=21) Two-factor solution Factor 1 National WB Economic stability How Australia is governed Business Wealth/income distribution Family support Economic situation Social conditions Health services State of environment National security Life in Australia Trust in people Life as a whole Happiness Standard of living Personal relationships Achievements in life Future security Comm. connectedness Health How safe you feel Eigen values Variance explained Total variance explained = 46.64% Comments: Factor 2 Personal WB .73 .71 .71 .70 .70 .69 .68 .67 .66 .61 .56 .39 .82 .81 .72 .71 .68 .63 .56 .51 .44 7.11 33.85 2.69 12.79 The five items comprising Factor 3 ‘social capital’ in the three-factor solution have been generally, if in some cases marginally, incorporated into the two major factors. Notably, however, Social Capital (Trust) actually fails to load on National Well-Being using a criterion of >.4. Note: Varimax produced an inferior solution, where health and trust both failed to load >.4. 5 1.4 Conclusions and Recommendations (a) The two indices have been generally confirmed. (b) In terms of the Personal Index, the factor strongly includes both ‘Life as a whole’ and ‘Happiness’ Recommendation 1: ‘Happiness’ be omitted from Survey #3. (c) In terms of the National Index, the factor clearly includes ‘Life in Australia’, but very marginally includes Social capital. The factor also fails to discriminate between the six ‘Index’ items and the four ‘specific’ items. Recommendation 2: That additional Social Capital item(s) be included in Survey #3 to replace 4 items in Factor 1. 6 2. Exploratory factor analyses for gender and event Due to the almost-emergence of a third factor of ‘ Social capital’ when all the items were combined (see 1.3.1), we ran a series of sub-sample analyses where the group was divided based on the impact of the USA attacks. Our rationale is that the emergence of this potential factor could be dependent on the extent to which people felt their sense of social capital needed to be drawn-on as a consequence of feeling threatened. We also explored whether the data factor differently between the genders. 2.1 Analyses based on event impact groupings The factor loadings (oblimin only) for the four items in question (Pattern matrix), N=3,000. Only the items loading onto the third factor are reported. Impact Group Not made sadder N= Item A. B. C. D. E. National security How safe you feel Future security Trust in people Connection to comm. Eigen values (whole factor) % variance explained Total variance explained Impact 4 or less N = 303 A. B. C. D. E. National security How safe you feel Future security Trust in people Connection to comm. Eigen values (whole factor) % variance explained Total variance explained Impact 5,6 N = 281 A. B. C. D. E. National security How safe you feel Future security Trust in people Connection to comm. Eigen values (whole factor) % variance explained Total variance explained Factor 1 National Factor 2 Personal .46 Factor 3 Social Capital .53 .56 .50 .49 .52 6.87 32.70 53.04% 3.17 15.09 1.10 5.25 .48 .58 .76 .36 .44 .63 2.43 11.56 1.31 6.26 .40 7.12 33.92 51.74% .70 .50 .56 7.81 37.18 54.98% .58 .58 -.32 2.65 12.63 1.09 5.17 7 Impact 7 N = 247 A. B. C. D. E. National security How safe you feel Future security Trust in people Connection to comm. Eigen values (whole factor) % variance explained Total variance explained Impact 8 N = 293 A. B. C. D. E. National security How safe you feel Future security Trust in people Connection to comm. Eigen values (whole factor) % variance explained Total variance explained Impact 9 N = 245 A. National security B. How safe you feel C. Future security D. Trust in people E. Connection to comm. (social conditions) (standard of living) (health) Eigen values (whole factor) % variance explained Total variance explained Impact 10 N = 455 A. B. C. D. E. National security How safe you feel Future security Trust in people Connection to comm. Eigen values (whole factor) % variance explained Total variance explained Conclusion: .47 .37 .40 .80 .34 .64 .33 2.86 13.60 1.20 5.73 .35 .86 .64 .60 .36 2.79 13.27 1.32 6.29 .35 .74 .50 .43 6.49 30.89 50.21% .68 6.97 33.20 52.56% .61 .30 .48 6.62 31.51 50.62 2.72 12.94 .45 .59 .35 .35 .31 1.29 6.16 .50 .66 .58 7.38 35.12 53.20% .50 .69 .42 2.71 12.90 1.09 5.18 (a) Only two groups (Impact 4 or less, and 8) produced a clean 3-item third factor. (b) Impact group 9 produced a quite different third factor that omitted ‘Trust’. (c) It is notable that ‘safety’ and ‘security’ items also comprise this putative third factor. This offers an alternative factor name as ‘security’. 8 2.2 Analyses based on gender groupings No additional factor patterns emerged using gender groupings. For both genders, the 3-factor solution failed to produce a distinct third factor, and the oblimin 2factor solution produced two distinct factors. The ordering of items within each factor also differed very little between the genders as shown below. Item Economic stability How Australia is governed Business Wealth/income distribution Family support Social conditions Health services Economic situation National security State of environment Life in Australia Trust in people Factor 1 Rank (loading) Male Female 1 (.76) 2 (.75) 3 (.72) 4 (.71) 5 (.70) 6 (.67) 7 (.67) 8 (.67) 9 (.64) 10 (.61) 11 (.46) 12 (.41) 1 (.74) 2 (.73) 4 (.73) 7 (.71) 3 (.73) 6 (.72) 5 (.72) 9 (.71) 10 (.62) 8 (.71) 11 (.61) 12 (.32) Happiness Life as a whole Standard of living Personal relationships Achievements Future security Comm. connectedness Health How safe you feel Eigen values Variance explained Total variance explained 7.09 33.75 46.21% Factor 2 Rank (loading) Male Female 7.18 34.17 47.14% 1 (.85) 2 (.84) 3 (.78) 4 (.78) 5 (.68) 6 (.60) 7 (.54) 8 (.52) 9 (.42) 2 (.86) 1 (.86) 4 (.69) 3 (.74) 5 (.69) 6 (.61) 7 (.54) 8 (.50) 9 (.43) 2.62 12.46 2.73 12.98 9 3. A comparison of Varimax vs. Oblimin on 2-factor solutions based on USA attacks impact Impact Group Variable Varimax Oblimin Not made sadder F1 Eigen Variance F2 Eigen Variance Total variance Clear factors¹ 6.87 32.70 3.17 15.09 47.79 No 6.87 32.70 3.17 15.09 47.79 Yes Impact = 10 F1 Eigen Variance F2 Eigen Variance Total variance Clear factors¹ 7.38 35.12 2.71 12.90 48.03 No Yes F1 Eigen Variance F2 Eigen Variance Total variance Clear factors¹ 7.12 33.92 2.43 11.56 45.48 No Yes Impact = 4 or less ¹ Loading criterion >.3 Conclusion: The oblimin rotation provides the best solution. This is also the preferred form of rotation as the two factors are significantly correlated with one another. 10 4. National items regressed against ‘How satisfied are you with life in Australia?’ This analysis was undertaken to determine the extent to which each Index item contributed to the prediction of the global item. 4.1 Standard regression (N = 2,004) 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. Economic situation State of environment Social conditions How Australia is governed Business National security Wealth/income distribution Health services Family support Economic stability p Zero order Unique %¹ .29 .05 .18 .02 .08 .07 .00 .03 .06 -.04 .000 .026 .000 .456 .002 .001 .941 .204 .014 .128 .485 .368 .425 .364 .396 .326 .333 .311 .340 .358 4.45 .17 1.99 .02 .34 .37 .00 .01 .20 .01 ¹ Part ² R = .562 R² = .316 Adjusted R² = .313 Unique variance = 7.56% Shared variance = 23.74% Conclusion: (a) The Index items generally performed better than the non-index items (b) The exceptions are ‘How Australia is governed’ and ‘Family support’ (c) The overall level of prediction is weak 4.2 Stepwise regression STEP 1. Economic situation (plus) 2. Social conditions (plus) 3. Business (plus) 4. Family support (plus) 5. National security (plus) 6. State of environment Adjusted R² R² change p .235 .236 .000 .293 .058 .000 .302 .010 .000 .307 .005 .000 .311 .004 .001 .313 .002 .016 11 Variables not appearing How Australia is governed Wealth/income distribution Health services Economic stability Conclusion: This confirms the greater predictive value of Family Support over How Australia is governed. Recommendation: That ‘Family support’ replace ‘How Australia is Governed’ in the National Index. 4.3 Hierarchical regression Step 1 National security State of environment Australia governed Economic situation Social conditions Business Step 2 Health services Wealth/income distribution Family support Economic stability Adjusted R² R² change p .310 .312 .000 .313 .004 .015 12 4.4 Hierarchical x Gender Adjusted R² R² change p Males Step 1 Step 2 .242 .244 .248 .006 .000 .255 Females Step 1 Step 2 .355 .356 .358 .003 .000 .148 Zero order Step 1 1. Economic situation 2. State of environment 3. Social conditions 4. Australia governed 5. Business 6. National security Step 2 1. 2. 3. 4. 5. 6. 7. Wealth distribution 8. Health services 9. Family support 10. Economic stability Unique (Part²) Male Female Male Female .42 .30 .40 .33 .34 .23 .52 .41 .44 .39 .44 .39 4.37 .23 3.57 .00 .26 .00 4.80 .30 1.88 .10 .45 .94 .36 .34 .36 .40 3.92 .13 2.53 .02 .26 .00 .00 .00 .38 .02 4.62 .20 1.64 .07 .40 .86 .00 .12 .08 .07 .30 .26 .31 .31 Conclusion: A great deal of similarity. 13 5. Index item intercorrelations 5.1 Personal Index item intercorrelations Domain 1 2 3 4 5 6 7 1. Life as a whole 2. Standard of living .62 3. Health .38 .38 4. Achievements in life .51 .48 .31 5. Personal relationships .51 .41 .26 .39 6. How safe you feel .28 .30 .24 .21 .20 7. Community connectedness .38 .35 .24 .33 .33 .33 8. Future security .44 .45 .29 .40 .33 .41 .40 B sr2 .36*** .34 .072 .08*** .09 .006 .19*** .18 .022 .21*** .23 .039 .02 .02 .000 .06*** .07 .004 .07*** .08 .004 ***p<.001 R = .720 R² = .519 Unique explained variance = 14.7 percent Shared explained variance = 37.2 percent 5.2 National Index domain intercorrelations Domain 1. 2. 3. 4. 5. 6. Economic Environment Social Governed Business National Security 1 .41 .43 .51 .55 .40 2 .51 .43 .41 .32 3 .45 .42 .34 4 .51 .36 5 .44 14 6. Personal Index item validation from Survey #1 Multiple regression was used to validate the domains of each Index. The seven domains of the Personal Index and the three domains of the National Index were regressed against the other variables. The results are presented in Table 2. 6.1 Prediction of other variables by the index domains Personal Index A ß sr² Dependent Variable B C D ß sr² ß sr² ß sr² E ß sr² Standard of living .32*** 6.4 .15*** 1.4 .06* 0.2 .13*** 1.0 .12*** 0.9 Health .14*** 1.5 .08** 0.5 .08*** 0.5 .10*** 0.7 .07** 0.3 Productivity .20*** 2.7 -.02 0.0 -.00 0.0 .04 0.0 .02 0.0 Relationships .19*** 2.9 -.01 0.0 -.01 0.0 .02 0.0 -.02 0.0 .02 0.0 .08** 0.5 .12*** 1.1 .06* 0.3 .06* 0.3 Community .08*** 0.5 .11*** 0.9 .15*** 1.6 -.02 0.0 .10*** 0.7 Future security .10*** 0.5 .11*** 0.6 .14*** 1.0 .16*** 1.5 .14*** 1.1 Safety R .73*** .38*** .40*** .46*** .32*** .52 .14 .16 .21 .10 Unique variability 14.5% 3.9% 4.4% 4.6% 3.5% Shared variability 37.7% 10.1% 11.6% 16.4% 6.5% Adjusted R² Key A = Life as a whole; B = Life in Australia; C = Social capital; D = Own life changing; E = Australia changing sr² = Percentage of unique variance Within the personal index, standard of living makes by far the largest unique contribution to the prediction of life as a whole. All other domains made a significant contribution of unique variance with the exception of safety. This domain also exhibited the lowest set of bi-variate correlates with the other domains (Table 5.1) and the weakest loading (.51) within the Personal Well-Being factor. On these grounds an argument could be mounted to exclude the domain of safety from the Personal Well-Being Index. This domain does, however, exhibit more relevance when used to predict other variables, while the contribution of other domains approaches zero. This changing pattern can be seen in columns B to E in Table 6.1. Here, for example, safety contributes 1.1% unique variance to the prediction of social capital (one quarter of the total unique variance), while the other domains of achievements and relationships 15 make no unique contribution to this prediction. A summary of these analyses, based on the Personal Index, domains, is as follows: 1. All of the variables (A to E) are significantly predicted by the combined domains. 2. The two domains that make no unique contribution to the prediction of variables B to E are achievement and relationships. Both of these, however, made a significant unique contribution to the prediction of ‘Life as a whole’ (Variable A). 3. The domains that provide the most consistent unique contribution to the prediction of variables A to E are Future Security, Health, and Standard of Living. Each made a significant unique contribution to the prediction of all six dependent variables. 7. Inter-item correlations Correlations N = 2,004 Future security Personal well-being index Life in Australia Economic situation State of environment .353** .464** .730** .270** .302** .241** .267** .248** .26 .239** .290** .589** .186** .204** .178** .172** .151** .21 .208** .334** .402** .665** .222** .289** .208** .177** .231** .22 1.000 .191** .329** .328** .641** .146** .142** .146** .138** .113** .13 .208** .191** 1.000 .346** .418** .588** .217** .215** .172** .201** .149** .21 .239** .334** .329** .346** 1.000 .403** .661** .206** .213** .266** .250** .220** .21 .464** .290** .402** .328** .418** .403** 1.000 .721** .265** .315** .251** .266** .249** .30 .730** .589** .665** .641** .588** .661** .721** 1.000 .328** .363** .318** .319** .294** .33 Life in Australia .270** .186** .222** .146** .217** .206** .265** .328** 1.000 .485** .368** .425** .364** .39 Economic situation .302** .204** .289** .142** .215** .213** .315** .363** .485** 1.000 .434** .425** .520** .55 State of the environment .241** .178** .208** .146** .172** .266** .251** .318** .368** .434** 1.000 .536** .451** .43 Social conditions .267** .172** .177** .138** .201** .250** .266** .319** .425** .425** .536** 1.000 .456** .41 How Australia is governed .248** .151** .231** .113** .149** .220** .249** .294** .364** .520** .451** .456** 1.000 .52 Business .261** .213** .220** .137** .210** .217** .307** .339** .396** .554** .434** .414** .526** 1.0 National security .200** .179** .159** .095** .251** .228** .260** .298** .329** .409** .343** .349** .371** .44 Wealth/ income distribution .317** .174** .227** .148** .201** .238** .320** .351** .333** .432** .431** .455** .500** .47 Health services .223** .114** .130** .096** .147** .145** .223** .234** .311** .343** .404** .414** .404** .40 Family support .264** .161** .199** .123** .166** .194** .248** .293** .340** .388** .435** .449** .425** .39 Economic stability .264** .175** .237** .110** .213** .225** .317** .333** .358** .589** .409** .424** .495** .52 National well-being index (new) .344** .247** .292** .174** .270** .316** .373** .437** .535** .755** .721** .711** .777** .76 Standard of living Health AchievePersonal How safe Comm. ments in life relationships you feel Connect. Standard of living 1.000 .395** .467** .421** .301** Health .395** 1.000 .308** .242** .236** Achievements in life .467** .308** 1.000 .385** Personal relationships .421** .242** .385** How safe you feel .301** .236** Community connectedness .353** Future security Personal wellbeing index How Social Australia is Busin conditions governed ** Correlation is significant at the 0.01 level (2-tailed). 16 17